Reminds me of Urbit sigils. They had a really cool system for generating unique visual identities. The write up is one of my favorite design/tech articles:
This makes me think of user-avatar scenarios, like systems that automatically make an abstract design linked to an anonymous user's IP.
If the possibility space is big enough, old heraldry rules could be applied the same way—though old families might not appreciate seeing "their" design randomly appropriated.
1. LLMs building tools that can be run by either LLMs or humans are a great idea
2. Generative art sometimes feels like "art by machine" but humans create art and logos in very similar ways e.g. start with "the rules" and then use that to create the art
Also, someone generally pops up in these discussions about "how this is not being creative" etc. To that I point people to the Everything Is A Remix series:
Here's an idea that I've retrained for the longest but never acted on:
Give people a survey, to identify movies, books, etc. they like. Then summarize all this into a visual symbol. Then you can print this on a t-shirt on a laptop sticker, etc. Humans get the general areas you like but it's machine readable, too - take a photo and you get all the details that he person chose to share!
Tangent, but I deeply appreciate the treatment of footnotes on this article, just floating alongside the text in the right gutter instead of at the bottom of the article. It seems so obvious (at least on desktop) but I haven't seen it before. It's so nice to be able to see whether it's something I want to read, or just a citation, without clicking and losing my place (yes, I know there's usually a link back to the article flow, but especially in the "sticky header" era we've been in for decades, that rarely works well).
On my tablet (always in landscape, not lacking space), with "desktop site" mode on, the sidenotes are turned into footnotes anyway. (The text is unnecessarily large: perhaps what happened was like mobile device? -> make text big -> no room for side notes now.)
Can we all agree to system prompt Claude and open models to stop using the same exact Fraunces font everywhere (The Capital J looks like someone folded it backwards). I said open models because qwen and Glm and Claude produce the same design and font for the same “design a gorgeous website” prompt.
I ran multiple experiments with models the past few weeks for generating designs and 80% of the time they use Fraunces if no selection is provided. Even if this was not AI generated it will appear as if it is because of this Font. It’s the new em-dash.
Reminds me of Urbit sigils. They had a really cool system for generating unique visual identities. The write up is one of my favorite design/tech articles:
https://urbit.org/blog/creating-sigils.
> Visual-Identity Generators
This makes me think of user-avatar scenarios, like systems that automatically make an abstract design linked to an anonymous user's IP.
If the possibility space is big enough, old heraldry rules could be applied the same way—though old families might not appreciate seeing "their" design randomly appropriated.
Related: Chernoff Faces [0]
In summary, humans are really good at noticing differences on faces so why not plot multi-dimensional charts on a face instead of a regular X/Y plot
0 - https://en.wikipedia.org/wiki/Chernoff_face
Quick visual on some mon:
https://mon.xavid.us/
https://en.wikipedia.org/wiki/Mon_(emblem)
Nice. I was craving more examples of this, as I'm not familiar with this tradition. "Cranes with Pine Branch" is my favorite on this list.
There was a really cool project a few years ago from Area Tech that explored this - generative shields (with heraldry on them). https://shields.build/
I love this for several reasons:
1. LLMs building tools that can be run by either LLMs or humans are a great idea
2. Generative art sometimes feels like "art by machine" but humans create art and logos in very similar ways e.g. start with "the rules" and then use that to create the art
Also, someone generally pops up in these discussions about "how this is not being creative" etc. To that I point people to the Everything Is A Remix series:
- https://www.youtube.com/watch?v=nJPERZDfyWc (2016)
- https://www.youtube.com/watch?v=X9RYuvPCQUA (updated for 2023)
Here's an idea that I've retrained for the longest but never acted on:
Give people a survey, to identify movies, books, etc. they like. Then summarize all this into a visual symbol. Then you can print this on a t-shirt on a laptop sticker, etc. Humans get the general areas you like but it's machine readable, too - take a photo and you get all the details that he person chose to share!
Thoughts?
Tangent, but I deeply appreciate the treatment of footnotes on this article, just floating alongside the text in the right gutter instead of at the bottom of the article. It seems so obvious (at least on desktop) but I haven't seen it before. It's so nice to be able to see whether it's something I want to read, or just a citation, without clicking and losing my place (yes, I know there's usually a link back to the article flow, but especially in the "sticky header" era we've been in for decades, that rarely works well).
Tufte CSS[1] provides a style for this, I like the good use of the "wasted" space in the margins of most browsers.
[1] https://edwardtufte.github.io/tufte-css/
Thanks! Also, if you squish the browser down to mobile-size, they turn into end notes. It took me a bit of experimentation to get it like this!
On my tablet (always in landscape, not lacking space), with "desktop site" mode on, the sidenotes are turned into footnotes anyway. (The text is unnecessarily large: perhaps what happened was like mobile device? -> make text big -> no room for side notes now.)
I was slightly disappointed of them still being labeled as "side notes" even if they were at the bottom
Heh. I guess one could have some fun with a ::before {content: "Side" or "Foot" } inside each breakpoint's media query.
Yeah, there is room for improvement there.
Can we all agree to system prompt Claude and open models to stop using the same exact Fraunces font everywhere (The Capital J looks like someone folded it backwards). I said open models because qwen and Glm and Claude produce the same design and font for the same “design a gorgeous website” prompt.
I chose this font myself, though. It was a human choice in this case. =/
This site is not LLM-generated.
I ran multiple experiments with models the past few weeks for generating designs and 80% of the time they use Fraunces if no selection is provided. Even if this was not AI generated it will appear as if it is because of this Font. It’s the new em-dash.