Post

What I want to point out, what models lack, is what we call intention, motivation, experience through countless failures. They have to derive what we want, what our goal (their mission) is from a few tokens we hand them in the prompt. Therefore the results are fairly impressive IMHO.

If you pay a lot yes, otherwise not. I had it for a couple of big app features/services that they lost context and broke their own previous code in the same session. But we’re not talking about normal websites here but complex services.

Can you site which LLM you have been using along with any other tools for this article? I find a few MCP servers have improved design to code outputs like figma MCP but the results massively vary per model and still not anywhere near 100% accurate.

based on when i experimented with claude my hunch was that their visual model turns images into a relatively limited number of tokens plus it does OCR to capture text. it was terrible at spatial relationships... which is exactly what you need for front-end stuff lol

It also sucks at making images 👀

30 years ago I said most over 40 would never grasp the power of the WorldWideWeb. 30 years later I say most over 30 will never harness the power of AI

fundamentally it has the same problem backend engineers (broadly speaking) have: trying to apply tech solutions to human problems. what we do is coding UX solutions and there is no debugger for those!

That feeling when I agree and disagree at the same time :) For a new, custom website with a creative approach, I agree with your post. However, if I use well-defined libs like MUI, which is well-known for AI, there are no issues at all. Yes, boring, but fast and perfect. Sometimes that is good enoug

Then for the design, maybe we could have models digging this way. I mean, dedicated models found pretty innovative CPU layouts! But LLM aren't those kind of models, they are just the wrong tool to get creative interfaces.

How does one define and communicate “taste” to an AI?

ai sucks at generating images that aren't repulsive and make one not read a post, too, btw.

imo this will probably always be true. there’s just too little good front end out there the bad stuff will still be bad the next training cycle. the new stuff will be a lot more bad because it will have come from AI. the good front end won’t surface to the top and be the dominant approach