TL;DR

Anthropic's Claude Design and Sivi AI look similar on the surface, but they generate two different things. Claude Design is Claude AI, an LLM, writing responsive HTML files. Sivi AI is an LDM (Large Design Model) generating free-form, layered, editable graphic designs. You can edit text in both, but only Sivi lets you move, resize, rotate, crop, and overlap elements on a pixel canvas. Claude Design belongs with Framer and Webflow for websites. Sivi belongs with Canva and Figma for ads, social posts, posters, and banners. The test that separates them is the second click: edit the text, then try to drag it.

—-

Last week I pitched Sivi to an investor. I said Sivi's LDM (Large Design Model) generates free-form, editable, on-brand designs in 72+ global languages.

She said, "Sona, Claude Design does that now". That's based on her recent experience creating a presentation with Anthropic's Claude Design. She said "I'll show it to you", and shared her screen to prove it. She clicked on the headline in the presentation and edited the text in place. "See? Editable. On-brand. Done."

I asked her to do one thing. Move the headline text two pixels up.

She couldn't. Not because she didn't know how, but because Claude Design's output doesn't work that way. That single drag is where the difference between an LLM generating HTML and an LDM generating a free-form design becomes visible.

This post is about that difference, why it matters, and what each tool is actually for.

What Claude Design actually generates

Claude Design generates responsive HTML files. It is actually Claude AI, a large language model, writing code with divs, sections, flex containers, utility classes, maybe a bit of inline styling. The browser renders the code into something that looks designed.

That code-first foundation creates a hard ceiling on what you can do with the output.

You cannot move elements. Every element sits inside a parent div in a document tree. There is no x and y coordinate to drag. Asking Claude Design to shift the headline 40px right means rewriting the HTML structure, not nudging a layer.

You cannot resize elements freely. Resize handles do not exist on a div the way they exist on a Figma frame. Width and height are properties of code, not direct manipulations on a canvas.

You cannot add new elements directly. There is no toolbar to drop in a shape, a text box, or an icon. To add anything, you go back to the prompt and ask Claude Design to regenerate with the new element included, which kicks off another agentic pass over the whole document.

You cannot crop or pan images. An image in HTML is an <img> tag with a source. There is no crop box, no pan handle, no focal point selector. If the AI placed the wrong part of the photo in view, you cannot drag inside the image to reframe it.

Fonts are locked to a small list. Claude Design ships with a fixed set of web-safe and Google Fonts. You cannot drop in a brand font from a foundry, upload a licensed typeface, or pick from thousands of weights and styles the way a real design tool allows.

It appears to have no vision over its own output. When you ask Claude Design to place an object inside a container, it crops most of the object hard against the container edges unless the object is present in center-middle of the image area. A designer would scale, center, and add breathing space around the subject. Claude Design clips. Image masks are limited to a small set of basic shapes (circle, rectangle, diamond, arch), not custom paths.

Tweaking is expensive. Every edit triggers an agentic flow that re-reasons over the document and regenerates code. Anthropic's published pricing for Claude Opus 4.7 is $5 per million input tokens and $25 per million output tokens, and a single design iteration in an agent loop can chew through thousands of tokens because the agent re-reads the full HTML context on every turn. Move a button, pay the tokens. Add an element, pay the tokens again. Compare that to dragging a layer in a canvas tool, which costs nothing.

What Sivi AI generates

Sivi AI generates a layered canvas using an LDM, a Large Design Model purpose-built for graphic design. Every element, text, shape, image, icon, background, sits on its own layer with x and y coordinates, rotation, opacity, blend mode, and z-index. The output is a design file, not an HTML document.

That means:

• Move any element anywhere on the canvas.

• Resize, rotate, overlap, group, lock, hide.

• Crop and pan images with a focal point.

• Use any brand font you upload.

• Mask with any shape, not just circles, rectangles, diamonds, and arches.

• Resize a 1080x1080 Instagram post to a 1080x1920 story and the LDM re-lays out the layers to fit the new ratio.

This is what an LDM is trained to produce. Not markup. Not a webpage. A free-form design composition with the same layout a human designer builds in Figma.

Divs vs pixels: the real dividing line

Claude Design belongs in the same family as Framer and Webflow. It is an interface design tool. The unit is the div, the layout is a document tree, and the output is a website. That is a legitimate, valuable category.

Sivi sits in a different category. The unit is the pixel on a canvas, the layout is free-form layered composition, and the output is a graphic. Ads, social posts, posters, banners, thumbnails.

Different substrates produce different ceilings. HTML is brilliant for responsive websites and terrible for a 1080x1080 ad where the offer needs to overlap the product photo at a 12 degree tilt. A canvas is brilliant for that ad and pointless for a scrolling marketing site.

Why the investor missed it

She clicked on text and edited it. That is the test most people run, because that is the test most tools pass. Notion, Canva, Figma, Webflow, Framer, Claude Design, all of them let you click and edit text.

The test that separates a document from a design is the second click. Grab the element and move it. If the layout reflows around your move, you have a document. If the element lands where you drop it, you have a design.

Claude Design fails the second click. Sivi passes it.

When to use which

Use Claude Design when you want a responsive website, a landing page, a structured HTML document, or a working frontend prototype. It is an interface design tool from Anthropic, and inside that lane it is genuinely useful.

Use Sivi AI when you want an ad, a social post, a poster, a banner, a thumbnail, anything that lives on a fixed canvas and needs free-form layered control with pixel-level precision.

Different models, different outputs, different jobs. The investor saw a product UI that looked like graphic design and assumed the underlying capability was graphic design. It was not. The moment you ask either tool to do the other one's job, the gap shows up in a single drag.