Yes, pretty much. The idea is to have a design file that both humans and agents can read and edit. Will it become a standard? Who knows, but I think that since there is nothing similar out there chances are high. And yes, it uses its own language, Elyx, with syntax that looks quite similar to CSS.

Design apps today often let you export HTML/CSS or React. But the generated code has no understanding of your production code architecture, so it needs to be reworked. You are also limited to the stacks the app supports. If you’re building a custom UI in C for example, a React export won’t help. And it’s a one-way handoff. You export the code, change it, and the design starts drifting from the implementation.

So .elyx file is like a two-way interface between designs and code, and a potential future open standard for designs.

You work on a visual canvas, but underneath the design is.elyx file. That’s the real magic here. It is a plain text file. A bit like CSS - describing components, layouts, tokens, etc.

Those files can live alongside your code. You can use git, review changes, compare diffs, use your own AI tools, or build tools around. Some people have already made tools to animate .elyx designs. A design project becomes much like any code project, you can organize and work with .elyx without Elyx app.

This makes it easier for everyone to work on a project. Designers, developers, and agents can collaborate much better. Designers edit visually, while developers and agents can read and edit the same files directly. Now, because those files describe the design without tying it to a particular framework you can use them to translate designs to whatever software stack you are using for your project.

When Max speaks, we listen! 🤘

@Danny the Cyber Guy says hi.

GPT Image 2 has had at least two major nerfs over the past month or so. I’m 100% sure of it.

I was producing a series of illustrations in a very distinctive, tightly controlled style. After the first nerf, I started using ChatGPT to explore scenes and characters, then doing the final renders through the API. That worked well for a week or two, but then they nerfed it even further in ChatGPT, now I'm not sure does API struggles to output same quality because reference images are so bad, or because it got nerfed as well.

GPT Image 2 has had at least two major nerfs over the past month or so. I’m 100% sure of it.

I was producing a series of illustrations in a very distinctive, tightly controlled style. After the first nerf, I started using ChatGPT to explore scenes and characters, then doing the final renders through the API. That worked well for a week or two, but then they nerfed it even further in ChatGPT, now I'm not sure does API struggles to output same quality because reference images are so bad, or because it got nerfed as well.

🔥🐷⚡️