I wonder if anyone saying "Claude Design is a Figma killer" has actually used CD in practice. I have, and CD is no different from asking an image generator to create a layout for you. It is a tool for quickly generating concepts, NOT a replacement. Just like code, we can't vibe our way to completing everything. Design needs a human in the loop to validate the fine points, unify components, and iterate to a final concept. Don't get me wrong, Claude Design is awesome - Anthropic is doing some cool stuff. But you can't directly manipulate flat concepts. You should not take what is given and just run with it - that isn't creativity, that is laziness.
I agree with the thesis. I'd say Google Stitch is more an analog to Figma if we're making comparisons. I feel like it's more of a human-in-the-loop/AI hybrid.
I think the entire point is that for many people in this world (non designers) it's a fast way to have all sorts of visuals. For designers and creatives - it's a starting point, ideation, exploration, template (whatever you call it) that needs further work or even rework but without a "blank canva" problem. Just like with Google's Stitch (I personally like it more, Claude's aesthetics is a bit too specific) - designs don't usually make it to production, but they do help in creative workflow.
As others are saying, it's just a tool. And I've found the best way to use it is in combination with other tools. It's great for brainstorming, taking care of the basics, prototyping, and its tight integration with Claude Code is helpful. I've had varying levels of success with Claude Code's output to Figma. html.to.design does a much better job, ime. Then you can do editing in Figma instead of burning tokens or credits attempting to tweak it in AI. Eventually, you can bring it back in from Figma to your AI of choice and go from there. Stitch is more basic. Lovable is more designer-friendly (but without a direct connection to Figma). A clear directive based in research, past experience, eye for design, depth of UX knowledge and decision making is what allows these tools to accelerate and elevate. It's more than just a human-in-the-loop scenario. Without that, you get yet more slop. UI Collective also has some good, free, recent videos about using these tools in combination.
Not only that, but what I’ve experienced is people asking Claude Design to create something that merely looks good, or makes a cool interaction model, and then calling it good and sending it to dev. But Claude Design isn’t going to ask you about your users. It isn’t going to ask you about adjacent systems to make sure that your design actually solves the need. It isn’t going to have the context to make sure that your design is consistent. It isn’t going to maintain a source of truth to keep the design consistent. And it’s damn sure not going to address the processes that can so easily destroy a good design solution if they’re not locked in. There’s absolutely a “design refining” process that has to happen between ideation and dev, and it often gets skipped because it takes time. AI gives you exactly what you ask for, but it can’t put in the thinking that, if you’re not a designer who’s spent years and years solving these problems, you aren’t experienced with.
Tools are not meant to replace each other. They're meant to be used best for different kind of work. You don't say that the hammer is a replacement of the axe, you use them both, but not at the same time.I do love using gsd skill for design that asks tons of questions about target audience and so much more before actually coding something. At least it's better than just typing into a void.
The problem is that there are so many bad designers who are hyping up these AI tools and claiming false results. Its like those bad designers can finally design with the help of AI. It creates such an anxiety in design field making you rethnk if what you are doing makes sense anymore. Good news - you have not missed a thing, Figma is still a go to tool for complex design systems and problem solving through UI. AI tools can speed things up by giving you many potential problem solutions in a short time.
Claude Design is a more versatile layer between founders/engineers and designers. Figma is too complex for non design folks. For various reasons, different orgs may or may not prefer design changes as a local branch, so instead they choose an environment where everyone can contribute in plain English. Claude Code still hasn’t figured out the collaboration layer not saying Claude Design has, but the entry point is much easier ish for some people
I think it's about some of those people aren't completely aware of the distinctions and how each tool differs. Figma is a superior design tool compared to Claude, but also Claude is not really trying to be a design tool. They already had the HTML/CSS generation in the webchat and API, might as well put a quick UI on it and make it a subproduct. You can't even move/reposition components in-grid lol.About the vibing part of it, I believe we're generally coming towards a tendency where the more efficient model is few super high skilled people using AI, at least how I currently like to think about it
I don't see what the hype is about Claude. It's broken every project I've attempted to set up, both established and new, and gives bad advice. I spent $40 for a month and I've already cancelled it. After 6 prompts in a premium model, it has the audacity to say I need a $140 upgrade. I've gotten way better results using a free ChatGPT account.
In my experience having tested a LOT of these tools (I have courses here on LInkedIn for a few of them), tools like Claud Design and Figma Make are GREAT for the first 80%, but SUCK for the last 20% (except that Figma Make is easy to bring into Figma Design). Sometimes 80% is good enough to test a new concept, etc. but sometimes it's not. Knowing which tool to use when is a key skill.