Been asked what's the difference between Brilliant & Paper and why either would be chosen over Figma. I'll start by saying Paper is an excellent tool and a great choice (not trying to be politically correct, but genuine appreciation for their craft and effort. Building a design tool is... extremely hard). Figma obviously set the standard. It's the most impressive web project ever built and a marvel of engineering. Seriously one of the most impressive engineering efforts in history. That said, here's the difference:show more

Paper: Agents generate HTML/CSS. The canvas is DOM-based. Therefore: high automation using AI (agents know HTML/CSS very well), limited manual editing functionality.

This isn't true only for Paper. Looking at Google Stitch, Claude design, Lovable, Base44 and every other HTML/CSS based canvas - it's very hard to have full professional depth when the underlying substrate is DOM-based. Not only because in some of the tools you're looking at a website, but there's a mathematical ceiling in each of these tools based on the tech stack they chose.

So HTML/CSS based canvases - incredible for agents, but have a hard time becoming a full blown editor.

Figma: Full manual control, professional depth. Everything needed to imagine and iterate on what a product should look like. But - proprietary file format, hard for agents to natively work on and reason about. Was also built starting 2012 without agents in mind.

Therefore, figma: High manual editing control, limited AI control

Brilliant: A professional vector design tool similar to Figma. Full manual control and the professional depth needed to imagine what a product will look like. Design systems, components, auto layout, vector networks, boolean ops etc.

Similar to Figma it has a highly performant GPU-native rendering engine and can handle real-world canvases of 100K+ elements.

But unlike Figma, Brilliant is built from scratch around Blueprint, a lossless language for design, tailored for agents (

Agents read and write blueprint, and from their POV it's like a super concise, highly expressive version of HTML/CSS (video below is not sped up)

It's actually more than that: The file format itself is blueprint (plain text .bl files) and even the rust wgpu rendering engine takes in blueprint which means you can have plain text blueprint design "blocks" in markdown. You can see an example of this here:

So Brilliant: Full manual control, professional depth AND high automation using AI

Also, the entire editor is free, unlimited MCP requests, and supports using your existing Anthropic, OpenAI, Cursor subs plus BYOK for any OpenAI-compatible AI provider. you can also try it, in the browser, without signing up here:

fully sold in :)

Objectively, it's amazing :)

This is very impressive... Love it!

Thank you! drop any feedback you have via cmd+k "Send Feedback" or just ask your agent

If performance matters (nice video), note this is lagging -

Thanks for this! Our agent fleet is already on this, trying to reproduce. Care sharing what browser/os you're using?

Latest Chrome / Mac OS.

thanks! please send anything else you see my way

Amazingg

YOU are amazing

<3

I agree with this the thing is is why I am going to use this if there is already something like @penpotapp that is literally infinity customizable and actually get updated for free, I understand the performance but actually agents are going to be moving 70% of the work in any design software. I just don't see, apart from convince, how is this adding to the scene in the long run, I actually like the product that you develop but I don't see it been adopted. I think your breakthrough on Dom base canvas and other things is amazing and I want you to win, but as someone that is working with Ai it's just crazy that I can open my ai harness tell it what I want and I can set up the use of penpot as a tool and just make insane presentations or anything pretty much, all local,self hosted and most importantly managed by the harness.

thank you! Penpot is also excellent and love their vision and thesis In case you missed it - Brilliant is (and will always be) Free. You can run it locally, offline, with your own AI sub to Antrhopic, OpenAI, and Cursor, or an API key to those plus Google, OpenRouter and any OpenAI-compatible provider. You can learn more about our thesis here if you're interested:

@penpotapp Amazing going to try it out and see if I can get brilliant to work on the workflows that I have already. Thank you for taking the time to make this clear. Will send feedback.

Interesting how the visual comparison turns implementation details into a clear design tradeoff.

👀

Just downloaded it. On the desktop version for windows, when hovering over Rectangle the "T" icon moves to the left and looses one pixel on the left! Super annoying! Other than that, awesome experience. Just told claude to recreate my web project. Did it!

Oh really! On it. Throw any other feedback you have my way

Yeah I DMd you.. But somehow can't send you screenshots there.. Probably you have to answer me first.

I've been using Brilliant, Pen (formerly Pencil), and paper in parallel to find my go-to agentic design tool. That said, my team and I still use Figma for documentation and as a source of truth. What I miss in all three products is a way to export designs back to Figma without friction. I understand from a business standpoint why they only offer import, but from a UX standpoint I think it's a miss. At the end of the day I'll still use the product for 99% of the work and pay for the subscription. Why lock the user into another ecosystem 🤷‍♂️ Btw I'm curious about your take on @pendev.

@usebrilliant Have you tried right click -> send to figma? NOT locking you in is kind of our thing :) As someone who's been using all three, I'd love to hear your thoughts and where you think we can improve