Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Scoring Show HN submissions for AI design patterns

adriankrebs.ch

Post reply on HN
Print view View original

221–230 of 258 posts

Re: Scoring Show HN submissions for AI design patterns

#221

Post by waterTanuki »

So it's true that AI-generated websites will follow similar style patterns given their training set and what the "modern" internet has converged on to be "optimal". But one could also argue based on the examples shown that any CSS template generated from shadcn's template generator (or any other for the most part) https://shadcnstudio.com/theme-generator also has a vibe-coded sheen to it.

Then the question becomes, do we need to go back to hand-picking every single css element to avoid being suspected of vibe coding? Why is it ok for someone to generate a css template on the fly using shadcn, but not ok to generate styles using claude code? Will someone using shadcn be judged the same as someone using claude code for styles?

Re: Scoring Show HN submissions for AI design patterns

#222

Post by CamperBob2 »

slibhb wrote: post #44
vunderba wrote: post #36

Earlier quoted context omitted.

My biggest issue with LLM‑assisted webpages (Claude Code is especially egregious) is the lack of respect for basic web content accessibility guidelines. The number of dark‑mode sites I’ve seen where the text (and subtext) are various shades of dark brown or beige is just awful. For reference, you want a contrast ratio between the text and background of at least ~4:1 to be on the safe side. This isn't even that hard t…

Just chiming in to say I don't care at all about accessibility and I find it bewildering that every thread sharing some project has a comment like this.

Just chiming in to say I don't care at all about accessibility

You will.

Re: Scoring Show HN submissions for AI design patterns

#223

Post by nozzlegear »

CrimsonRain wrote: post #80
vunderba wrote: post #36

Earlier quoted context omitted.

My biggest issue with LLM‑assisted webpages (Claude Code is especially egregious) is the lack of respect for basic web content accessibility guidelines. The number of dark‑mode sites I’ve seen where the text (and subtext) are various shades of dark brown or beige is just awful. For reference, you want a contrast ratio between the text and background of at least ~4:1 to be on the safe side. This isn't even that hard t…

I think accessibility is a really admirable thing and helpful to society (like ramps or parking). But stop shoving your wants on others when you can fix it on your own. Just write a chrome plugin using ai that adjusts css to set contrast ratio of your choice. Can even use a local llm to figure out replacement colors. Accessibility that can be had on client side should not be a concern on server side.

Our only recourse is to "stop shoving our wants on others," and instead make our own shit sandwich to fix someone else's shit sandwich?

Re: Scoring Show HN submissions for AI design patterns

#224

Post by stingraycharles »

sen wrote: post #219
sowbug wrote: post #136

Earlier quoted context omitted.

Were/Are human-generated side projects better in this respect?

Yes, because a majority just use a CSS framework to save time, and all the big/common frameworks have put a fair bit of effort into their default colours and typography. I would rather go back to when all side projects used Bootstrap than this purple-on-purple-with-glowing-purple mess of stuff we have now.

It’s also interesting that you can tell which LLM provider was used to build certain sites, sometimes even model and version.

It’s insane though how many dark mode websites with purple there are right now, how come all LLMs converged on that style?

Re: Scoring Show HN submissions for AI design patterns

#225

Post by vunderba »

conception wrote: post #220
vunderba wrote: post #84

Earlier quoted context omitted.

I could see that. I’ve found that the more specificity you add to your prompt and less freedom you give Claude Code to kind of just “do its own thing”, the better your results will be. FWIW, there’s also an official frontend-design skill for CC [1]. A while back I incorporated some of the more relevant guidance from WCAG into it [2]. [1] - https://github.com/anthropics/claude-code/blob/main/plugins/... [2] - https://…

Note guidance for 4.7 specifically calls out being more specific compared to 4.6. Though the system prompt seems to say the opposite. Tldr I would expect different outcomes with 4.7 with not being specific.

Honestly I'd heard so much seemingly conflicting information about the quality of 4.7 that I've got an override in CC to stay on Opus 4.6 (1 million token context) for the time being.

For reference, in the .claude/settings.local.json

  {
    "model": "claude-opus-4-6[1M]"
  }

Re: Scoring Show HN submissions for AI design patterns

#227

Post by p1necone »

postalcoder wrote: post #81
slibhb wrote: post #44

Earlier quoted context omitted.

Just chiming in to say I don't care at all about accessibility and I find it bewildering that every thread sharing some project has a comment like this.

Accessibility is a broad umbrella of features that enable a ton of really cool stuff for everybody, not just the disabled. Things like agentic computer use is only possible because of "accessibility".

The same accessibility stuff that makes screen readers work well also makes automated UI tests simpler and less brittle too (correct aria roles, accessible names, label relationships etc).

Re: Scoring Show HN submissions for AI design patterns

#228

Post by aetherspawn »

The problem is the bar of expectation has really raised since AI, now you absolutely must have a fancy website with 3-dozen pages and SaaS-like styling.

Before, you could get away doing business with a basic 1-pager, which is about the same as what everyone else had, but these days looks lazy/incompetent.

You don’t have any more time to throw it together than you did before so… yeah I guess slop it is. Probably not going to be humans reading it past the front page anyway. If you want to engage humans, use LinkedIn or TikTok or something.

Re: Scoring Show HN submissions for AI design patterns

#229

Post by Der_Einzige »

doug_durham wrote: post #39
mccoyb wrote: post #23

The problem is not vibe coding itself. The problem is that certain untrained people do not have or perhaps do not care to learn the necessary skills to refine the result into something novel, or clear / precise, something which communicates (clearly) the idea they are trying to convey to others (who are hoping to learn something new). In a climate where it seems like VC are woefully bereft of the same skills, there's…

Why is that a problem? Reality will filter out the projects that are poorly developed just like it always has.

Like it filters windows OS for making the start button an electron app? Oh wait, it didn’t and nothing is displacing windows.

Reality doesn’t filter out shit man. We objectively had better vaccumes in the 80s (central vacuums were popular then). We had objectively better displays in the 80s too (vector displays/game consoles like the vectrex were putting out variable refresh rates into the 1000hz equivalent with extreme brightness and impossible to replicate phosphor glow.

Worse is better. Like, actually getting worse is progress. Sometimes it really is good to “return to monkie” and take a page from the past.

Re: Scoring Show HN submissions for AI design patterns

#230

Post by joegibbs »

michaelcampbell wrote: post #15

> A designer recently told me that “colored left borders are almost as reliable a sign of AI-generated design as em-dashes for text”, so I started to notice them on many pages. so, n=1 plus Baader-Meinhof? ( https://en.wikipedia.org/wiki/Frequency_illusion )

AI definitely does seem to want to add coloured left borders, tags and superfluous numbers all over the place from my experience, you have to tell it specifically not to
Post reply on HN