> As another example, before AI what were the connotations of the sparkle emoji ? Personally, I don’t know, but now it means AI.
> For example, a lot of AI apps use tiny icons
> There are other aesthetics my brain associates with AI, like beige/cream colors, orange accents, and serif typefaces as well as whack-a-mole UI controls
Are examples of the AI outputs
> For example: streaming text
> For example, the “shimmering text”
Are examples of interacting with AI
But how is AI outputing sparkle emoji or orange accents colors interacting with AI? Unless you also classify AI output as a form of interacting with AI
The beige UI with orange accents probably refers to the design guidelines of Anthropic, Mistral, Microsoft AI and a bunch of other companies. They all try to use soft, non-threatening colours, and for some reason those two are a popular choice.
- https://www.salesforce.com/artificial-intelligence/
- https://copilot.cloud.microsoft/
- https://upload.wikimedia.org/wikipedia/commons/7/72/Meta_AI_...
- Literally the Gemini logo
I’ve never seen or heard anyone mention sparkle emoji as a tell-tail sign of AI generated content, in the way that you hear about em-dash for example.
But I have seen the icon used in just about every website/app that wants to let you know they are using “AI” in some form or another.
And that contrasts to an extreme degree against the rest of the scene which flows extremely well, maintains perfectly color coordination, and so on. I think it's kind of like how self driving cars can drive phenomenally well but then have occasional failure modes like randomly driving head first into a concrete divider. It's not like that's bad driving in the sense of missing a nuance, but something that's just completely conspicuously outside the rest of its demonstrated abilities.
That said, the provided examples definitely are a clear step up from the "default" you'd get. Like how some bootstrap-templated sites did it much better than others, while the best made it nearly invisible.
If we think about that as the intended competition (and not websites corps spent millions on design/polish/copy), I'd argue that AI (model + coding agent + skills + whatever) is doing a better job overall. Most agencies don't spend enough effort in understanding what you want, or hire juniors and bill them as seniors. It's almost always late, overbilled, somewhat what you said you wanted but not really, and technically terrible.
The Neon Ramen example look really cool at first glance, but if you spend more than a second or 5 looking at it you keep finding visual glitches, inconsistent design, and elements which don't belong. It's nice for a brainstorming phase, but that's about it.
If you were to go with a buy-a-template you'd probably end up with a less flashy design, but at least it'd be consistent. Those templates are built by combining standard styled elements to create a bunch of common design patterns. Sure, nothing looks great, but nothing looks bad either!
If you don't really care about a website, going for a ready-made template is by far the safer, easier, and cheaper option. And if you go for something like Squarespace, you don't even need an agency at all.
There's a lot of detail, but most of it is decoration for the sake of distraction.
AI isn't good at understatement, subtext, or (ironically) elegant distillation.
Is stylistic ingenuity, breaking from convention, or leaving things out on a website for a Ramen restaurant chain a feature or a bug? I'd choose to go based on the quality of the food not how unusual the website UI is. The website is not the product.
I personally find navigating and ordering food on restaurant websites that don't use standard conventions a frustrating experience.
This summer I've noticed a lot of menus with AI generated images of food and I have not trusted a single one of them. If you are gonna pitch me with this thing that blatantly is not even tangentially related to your actual product, that does not give me a lot of confidence in your business. If they'd used a stock image of a carbonara I would be none the wiser, but the entirely fake stuff is a massive turnoff.
Some of the best restaurants I know have old timey bare bones websites because they've been around for decades, and it's an oddly good predictor of a place that has a lot of loyal customers.
If I saw a website that looked like the ones above, I would be immediately skeptical. Handmade but rudimentary says down-to-business no nonsense, shiny and nicely made by a pro says high effort. This stuff? It says slop shop.
The thing that really gets me about this stuff is - is all about the content, the actual content is what really matters, you need to have content that actually works, that people find appealing, otherwise you have dogshit. A staccato bullet point of marketing speak is dog shit. It’s nothing, it’s less than nothing, customers are repulsed by it, and rightly so. bad presentation you can sometimes get away with, but bad content? Almost never. And this is bad (LLM generated) content, at a glance.
And now it's tweetable!
seems AI to me. Humans were doing marketing slop this way before AI but it doesn't really matter if the slop is AI or human.
I literally just got my SOC2 compliance complete an hour ago, and I'm gonna roll out team plans/enterprise plans next week (unified billing + shared brands). For teams I'm gonna charge a flat 25% margin on API costs, and for enterprise I'll have a 50% margin + minimum spend.
I think I'm going to keep the individual plans at 0% though. Realistically until my own diffusion models I'm training are the main model I'm using for generation, I'm just an API wrapper + harness. At the individual level I want to be competitive with going directly to the api and coding your own harness.
I am curious though - what would you pay per image?
If I’m using this towards something that makes me money, I’d pay a lot more, but it really depends on context. In general, I’ll spend $5 like it’s nothing, $20 if it’s nothing and it’s for a business expense, and probably $50-100 for a design scheme for my (quite small) business. I don’t know that I’m a representative sample, but I’ll put my money where my mouth is, because I just added an extra $5 to play around more.
Btw, it’s a really innovative interface. It’s fun and very different from other tools I’ve seen. I quite like it but it’s also a bit confusing. Feel free to contact me if you want customer feedback (info in profile).
https://en.wikipedia.org/wiki/Cost-plus_pricing / https://www.costplusdrugs.com
I know they probably intend it as "[Cost-plus] drugs", but I can't help but read it as "I'll do the job for you for the material cost, plus some drugs".
But “Drugs” wouldn’t be perfect…
Anyway no but now that you mention it maybe it’s too wonky of a name. Nothing better coming to the top of my head right now.
Sometimes models are vastly underpriced. Fable is an example. I can easily spend hundreds of dollars using it to code. On the other hand, if I'm looking for AI to review a contract for any obvious mistakes, I don't really care if I spend $0.01 vs $1.00, even if the difference is only marginal. I'll pay quite a bit more for something that's just a little bit better.
Same thing with design, in many cases. When the absolute dollar amounts are small, I'm happy to pay more for even marginally better quality.
You definitely can do fun/creative things with LLMs, but they need a lot more steering/handholding. IE if you add an additional page to the lazer kiwi leaderboard, it's likely the first pass will be somewhat more generic than the homepage.
One of the other advantages of diffusion is that it does really, really well with style transfer, meaning designing additional pages in the same style ends up pretty easy. As an example, here are designs of a user profile page and a top guilds page, using the look/feel of a screenshot of your homepage: https://image.non.io/5415e0e0-9264-4bad-988e-8d9d3885b518.we...
Stuff is jumping around after I try to move the canvas around, unexpected stuff. I can’t figure it out.
On the Tarot one, the 3D card in the bottom is both very elaborate and also pointless. No human would've put effort into making that. Not bad per se! But AI.
The reason I feed in the json blob is it lets me better preserve subjects across multiple screens (ie you probably want to carry over the header exactly as it is in one image to another image).
The example sites above though were pretty simple. I think they were something like
"A tarot card website with an inky, painted artistic style, with rich illustrations on a card featured on the left, and a button to draw cards on the right"
and for the other it was something like,
"A cyberpunk inspired ramen cart website"
But basically diffui's build step provides an API to create normal/roughness/depth maps from images. It's essentially an interface to fal.ai's Patina model, which does a really good job at creating those.
The one thing that's missing with it though is metalness maps (Patina can theoretically output them, but they only have a ~10% success rate I've found). I'm trying to train my own diffusion model now to help generate those so the reflectivity isn't fully uniform. I just posted very, VERY early results here - right now my model has around a ~25% succes rate: https://x.com/pwnies/status/2082980850120163756
The funny thing is that looking like everyone else was once considered the definition of good UI design, because it helps users understand a new interface. Then mobile apps came and suddenly UI design was about “expressing your brand” or whatever.
I actually think the other things that Jim mentioned, such as the async task gradient / standard sidebar for agents are good things.
That was really a test of whether or not newer models could implement box shapes that aren't provided by CSS. Opus 5 was one of the first that really nailed that - it's been a long standing test I've had for img->html.
If you could access the current models as base models I guess you'd still get very interesting hilarious stuff.
You are absolutely right, that's the load bearing assumption and it earns its keep.
Personally, I didn't mind when a lot of sites looked like the default Bootstrap theme (because that's what they used), because generally it made sense. I think that LLMs, both in regards to prose and the code and UIs they produce, could have lead to an age of decent - where things would be just fine and sane, the writing decent and not frustrating, the code easy to read and debug, UI whitespace and layouts would be good, fonts would be readable, usability would be great.
Instead we got AI slop writing that reads like someone turned the pop-sci/LinkedIn tone dial up to 11 and UIs that all feel like some sort of SaaS landing page.
Oh well, it could have been worse.
I think this is more due to temperature (while the LLM is producing a design) rather than consistency. Consistency means it'll be yes-and-ing all the way once it starts, but that initial seed? You meet one instance of Claude 5, you've met them all. Ditto ChatGPT-x or Gemini-y.
Because each blank context window starts the same, the only initial seed of variation is the user prompt. If the users don't mention style, then 5000 people all prompting for "single page web app to do X" will get very close to the same result.
Most humans keep asking for the same, and the output is therefore analogous to Simon Willison's pelican on a bike, where the output is overwhelmingly facing to our right, and to the extent that the bike is functional and would actually e.g. steer, it's a standard bike rather than recumbent, cargo, or tandem.
Something to consider regarding the narrow space AI-created designs align on: LLMs are trained to write consistent code. This makes sense for something like a billing or a backend function - you want that code to be consistent. The problem though is that LLMs write code to represent designs as well, which means you get consistent designs. You end up aligning on a generic mean because of it, which is often why you see these repeated aesthetics.
It was something I saw when I was working on AI tooling over at Figma. It was very hard to get creative, unique outputs out of LLMs.
One thing I'd recommend if you're trying to avoid this is try a diffusion model as a starting point. Gpt-image-2 is a VERY capable designer, and Opus and Fable are fantastic at converting images to webpages. Starting with images will let you sidestep a lot of the uniformity of output that LLMs have. I'm heavily biased here as this is what I left figma to build (https://news.ycombinator.com/item?id=48995754), but even starting with gpt-image-2 to give you a general sense of the look/feel will heavily differentiate you on the design front.
Here are some examples of image->webpage outputs I've been playing with:
https://html.non.io/tarot/
https://html.non.io/neonRamen/