claude design

me: help me design a page that shows token usage over time claude: *hardcodes it to 100%*

"Claude can help with many things, but finding this page isn’t one of them."

Just finished learnjng UX UI at a university and got a certificate a month ago and now this is a huge step for UX UI designers this is amazing cant wait to test it out

Amazing, now we can design a button every 5 hours

Sorry, I hit my session limit by just thinking how I'd use that.

Not shown: usage limit being cooked at the 0:20 mark.

can i use this on existing project

Your tutorials are awesome and really easy to follow! However, I’ve noticed that most people doing 'vibe coding' only stick to single-page sites. I’d love to see someone go deeper into MPAs (Multi-Page Applications). I'm curious how the workflow, prompt structure, and logic change when building something more complex—like a full-stack site with a front-end, back-end, database integration, and e-commerce features. A tutorial on that would be incredible!

This is all awesome!!!!! 🔥 But in reality, we already have active projects in progress using Lovable, Cursor, and Claude Code. Could you create a video showing best practices for applying Claude Design to real, in-progress projects? Specifically, I’d love to see: • How you approach an existing codebase (not starting from scratch) • The exact prompts you use to guide Claude Design • How you generate or evolve a design system from what’s already built ❤️ • How everything ties back into tools like Lovable and Cursor in a real workflow That would be incredibly helpful 🙌

best login method for it

This is valuable. Can you make a video showing how to build application with it.

I wish you had made it more clearer and complete for Total Beginner from A to Z (without assuming Prior Knowledge on Claude ai...). it is just you have assumed all your audience have prior knowledge and started from some where in the middle

Your tutorials are awesome and really easy to follow! However, I’ve noticed that most people doing 'vibe coding' only stick to single-page sites. I’d love to see someone go deeper into MPAs (Multi-Page Applications). I'm curious how the workflow, prompt structure, and logic change when building something more complex—like a full-stack site with a front-end, back-end, database integration, and e-commerce features. A tutorial on that would be incredible!

Hi Nate, thank you for this Presentation. would be interesint if you could make a more detailed tutoril about claude design, where you create a website or an app from scratch! :)

This is all awesome!!!!! 🔥 But in reality, we already have active projects in progress using Lovable, Cursor, and Claude Code. Could you create a video showing best practices for applying Claude Design to real, in-progress projects? Specifically, I’d love to see: • How you approach an existing codebase (not starting from scratch) • The exact prompts you use to guide Claude Design • How you generate or evolve a design system from what’s already built ❤️ • How everything ties back into tools like Lovable and Cursor in a real workflow That would be incredibly helpful 🙌

me: help me design a page that shows token usage over time claude: *hardcodes it to 100%*

Claude after asking it to help design an app: "Sure, but I'm gonna need to see your government issued ID first."

This is all awesome!!!!! 🔥 But in reality, we already have active projects in progress using Lovable, Cursor, and Claude Code. Could you create a video showing best practices for applying Claude Design to real, in-progress projects? Specifically, I’d love to see: • How you approach an existing codebase (not starting from scratch) • The exact prompts you use to guide Claude Design • How you generate or evolve a design system from what’s already built ❤️ • How everything ties back into tools like Lovable and Cursor in a real workflow That would be incredibly helpful 🙌

Thank you, Nate. That was an interesting and nicely structured video! Not stopping at the local website but how to push it out, and - to some degree - even edit it. I like this one!

coming with that heat! thanks for the tutorial... pretty mind blowing what you can do when you combine these AI tools (e.g. coding, image, and video tools) in a workflow like you showed

Highlights by TinaMind: 00:00 Claude Design, powered by the Opus 4.7 vision model, allows users to import brand assets and design systems to create consistent prototypes, slides, and web pages. 02:37 The platform includes features for organization-scoped sharing and direct hand-off capabilities to Claude Code for building, syncing, and deploying projects to GitHub. 06:16 Users can generate professional slide decks by providing PDFs or transcripts, utilizing the established design system to ensure brand consistency across all slides. 08:24 The interface offers interactive editing and feedback tools, including commenting, direct tweaks to layout elements, and real-time presentation modes. 10:00 Claude Design simplifies the landing page creation process by asking specific questions to guide the content and allowing users to customize elements like countdowns and CTAs. 12:21 The integration between Claude Design and Claude Code creates a cohesive workflow, allowing users to prototype in a visual environment before handing off to code for deployment.

This model is honestly insane. That GTA-style clone had a level of detail I haven’t really seen before, especially without much instruction. Just something to keep in mind though, they may be training for these demos. It might be worth switching up your tests over time and making them more dynamic. If they know the kinds of tasks reviewers usually run, the model could be optimized for those exact scenarios. In that case, it might look incredible in those demos but be less consistent in real-world use. Rotating tasks, adding new challenges, and avoiding repeatable patterns would give a much clearer picture of what it can actually do.

Wild how far Opus 4.7 has jumped in one version. The fact it can spin up full OS‑style UIs, 3D scenes, and working games from single prompts is honestly more impressive than most benchmarks. This feels like the first glimpse of ‘AI as junior dev / designer’ actually being real.

I keep coming to you for all the newest models. Great video!. I just hope anthropic won't nerf it in a few months :(

Aweome Nate. This have visual impact, but is not a websiet, this are landing pages. You have a big gap between both wordls regarding usability and architecture. Cheers!

Cool Video, gonne try it!! How do you do good SEO when building a website like this and how do you put contact forms on it??

😂 Darn, he didn’t kill anything. This type of design is comparable to work done by a junior-level designer.

The OpenRouter free model setup is clever, but worth flagging for non-devs watching: the free tier on most providers means your prompts and outputs are used for training. That's fine for personal projects but a real issue if you're building anything with client data, internal company info, or anything you'd consider sensitive. Worth checking each model's data policy before using it for work projects. I'm a PM with 20+ years in cybersecurity, building products with Claude Code. Privacy trips up a lot of non-devs because nobody warns them

Nate how do you deal with this when say you are doing consulting on client data, is there a way to ensure that this data isn't share with anthropic, similar to how copilot has a 'work mode'

A couple extra tips from someone doing this a lot lately - 1. Use Cowork 2. You can create your own skill with client-first info and extra instructions - for example I rely on variables heavily so you can get it to create them for font sizing - spacing etc. 3. It will not deal with components so what I do is tell it explicitly to add the suffix _component to repeated sections and containers so I can quickly spot them 4. If you start with the Relume cloneable you can include in your Skill a note "refer to your style guide for base styles and adjust them as needed to fit the Figma system" or visa versa 5. It is possible to get it to match the images for you if you pre-upload them to the media library and give them a descriptive file name in advance 6. Even in vibe coding apps, when it comes to recreation the best results I've seen do come from Figma connections, by far compared to screenshots or instructions. Indesign files somehow also work perfectly. I also recently tried the native Ai builder - with Claude as a coworker to make corrections. It's come a long way, not saying it's comparable at all but for a budget, rush job, that's was lucrative for me as well.