惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

GbyAI
GbyAI
Martin Fowler
Martin Fowler
I
InfoQ
腾讯CDC
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
爱范儿
爱范儿
Microsoft Security Blog
Microsoft Security Blog
Google DeepMind News
Google DeepMind News
D
DataBreaches.Net
云风的 BLOG
云风的 BLOG
F
Fortinet All Blogs
N
Netflix TechBlog - Medium
博客园 - 聂微东
Microsoft Azure Blog
Microsoft Azure Blog
D
Docker
博客园 - 三生石上(FineUI控件)
Y
Y Combinator Blog
博客园 - Franky
Engineering at Meta
Engineering at Meta
B
Blog
罗磊的独立博客
Apple Machine Learning Research
Apple Machine Learning Research
Jina AI
Jina AI
V
Visual Studio Blog

Hacker News - Newest: "AI"

AI can't read an investor deck AI as an attorney? Student uses ChatGPT, Gemini to sue UW over alleged racial discrimination Hacking MCP Servers in AI Systems – The Rug Pull: Tool Changes After Approval GitHub - MeepCastana/KubeezCut: Free Web based video editor Can AI judge journalism? A Thiel-backed startup says yes, even if it risks chilling whistleblowers Coming soon: 10 Things That Matter in AI Right Now DARPA built an AI to fact-check enemy weapons claims What explains heterogeneity in AI adoption? When AI Meets Muscle: Context-Aware Electrical Stimulation Promises a New Way to Guide Human Movements - Department of Computer Science AI Changed How We Build. It Did Not Change What Matters. Linux rules on using AI-generated code - Copilot is OK, but humans must take 'full responsibility for the… Meta spins up AI version of Mark Zuckerberg to engage with employees Code Mode: Let Your AI Write Programs, Not Just Call Tools | TanStack Blog GitHub - Delavalom/graft: Go framework for building AI agents. Type-safe tools, multi-provider (OpenAI, Anthropic, Gemini, Bedrock), zero vendor SDKs. India's TCS tops estimates, says new AI models did not dent services demand Gen Z's fading AI hype Strong feeling: we are in a folded AI reality GitHub - machinarii/total-recall-catalog: A reference catalog of latest knowledge retrieval, memory & RAG systems GitHub - mensfeld/code-on-incus: Give each AI agent its own isolated machine with root, Docker, and systemd. Active defense detects and stops threats automatically.. Quantization, LoRA, and the 8% Problem: Benchmarking Local LLMs for Production AI Iran war: We spoke to the man making Lego-style AI videos that experts say are powerful propaganda Powell, Bessent discussed Anthropic's Mythos AI cyber threat with major U.S. banks GitHub - immartian/bellamem: Persistent belief-graph memory for AI agents. Retrieves decisive context by importance — not recency, not RAG, not /compact. recursive-mode: The Repo-Native Operating System for AI Engineering After the attack on Sam Altman's home, will AI CEO's go on the offensive? The biggest advance in AI since the LLM Opus 4.6 vs GPT 5.4 One Prompt Unity World Generation Test “AI polls” are fake polls Client Challenge Can AI be a 'child of God'? Inside Anthropic's meeting with Christian leaders
AI Is Good at Web Design Now - Repaint
Ben Shumaker · 2026-06-16 · via Hacker News - Newest: "AI"

For the last few years, most websites made with AI have been generic and mediocre. That's why so much of it gets called slop. But in the past couple of months, I've noticed a major shift that nobody seems to be talking about: AI has gotten dramatically better at web design.

I've got a better view of this than most people because I'm the ceo of Repaint, an AI website builder. At this point, I've looked at thousands of websites people made with AI, and lately the good ones have gotten noticeably better. Here are a few that turned my head:

Homepage for a Singapore Kyokushin karate dojo with a dark hero, bold serif headline reading "Passion. Patience. Perseverance." and a photo of a karateka in stance

Homepage for a performance neuro training coach with a bright "Optimize your energy. Reset the system." headline, neon green highlights, and a photo of Dr. Paige Roberts

Homepage for a Belgian event crew company with a black and yellow layout, a "Sterke crew. Elk event." headline, and a photo of a crew setting up an arena stage

Pretty good, no? I certainly wouldn't call them slop. I'm pretty sure these are better than what most web designers would come up with.

What impressed me most was the aesthetic. Six months ago, AI models would never make websites that looked this unique. The AI simply didn't have enough taste. Before, you had to provide all the taste yourself with precise instructions to get a site like this.

But now the AI can get much further. Each of these sites still took a few days of iteration, but they had 80% of the final aesthetic within the first few versions. And they were made by business owners just talking to AI, not professional agencies.

So what changed?

I wish the secret sauce was Repaint, but it was mostly the models. We use a fairly standard AI agent with models from Anthropic, OpenAI, and Google. We've built special tools for rebuilding existing websites, but nothing groundbreaking that would explain this jump in design quality. Any tool using the latest models has probably seen a similar improvement.

Right now, our main model for website design is Opus 4.8. Anthropic released several updates in quick succession:

  • Opus 4.6 - February 5, 2026
  • Opus 4.7 - April 16, 2026
  • Opus 4.8 - May 28, 2026

I started noticing the shift in early May, though it may have already been there when we were using Opus 4.6.

In our testing, models from OpenAI and Google have not improved at web design to the same degree. GPT-5.5 is smart, but it makes noticeably worse websites. If you're an OpenAI employee reading this, please make the next one better at websites.

The Difference

The biggest difference is that the new models are much better at following design direction. To show what I mean, I had two agents build the same landing page from the same prompt. The first used Opus 4.5, an older model. The second used Opus 4.8, a newer model. Both had access to an image generation tool. Here's the prompt:

Create a website for an architecture firm. It should be inspired by Bauhaus design, with a minimalist aesthetic. Mostly neutrals, with a few bold colors. It should use broad images that fill lots of space.

Older Model - Opus 4.5

Hero section made by Opus 4.5

Featured work section made by Opus 4.5

This site looks polished at first glance, but almost all of that comes from the images. The design itself is just a simple layout with big photos. It ignored the bold colors and never developed a recognizable Bauhaus aesthetic. That was the limitation of older models: they could make something clean, but they couldn't really translate a design direction.

Newer Model - Opus 4.8

Hero section made by Opus 4.8

Featured work section by Opus 4.8

I'm not sure this site is objectively better, but it followed my directions much more closely. Opus 4.8 found clear ways to use bold colors and Bauhaus-inspired design throughout the site. The result feels genuinely distinctive, and it is obviously based on the prompt.

The new models can follow instructions across a much wider range of styles. You can google "web design styles" and find a million directions to try. Change the visual direction in the prompt, and the entire website changes with it.

These are all one-shot generations:

Brutalism

Create a website for an architecture firm. It should feel inspired by brutalism, with a raw and heavy look. Use mostly concrete tones, black, and white, with one or two sharp accent colors. The layout can feel a little rigid and oversized, with broad images that take up a lot of space.

A brutalist architecture firm site

Playful Colors

Create a website for an architecture firm. It should be inspired by playful postmodern design, with a bright and energetic look. Use bold color combinations, simple geometric shapes, and a mix of clean and quirky details. The layout should still feel polished, with broad images that take up a lot of space.

A playful postmodern architecture firm site

Retro Futurism

Create a website for an architecture firm. It should feel inspired by retro futurism, with a bold and slightly cinematic look. Use dark neutrals, metallic tones, and a few bright electric colors. The layout can feel more experimental, with broad images that take up a lot of space.

A retro futurist architecture firm site

A year ago, no matter what you asked for, AI would make something like this:

A generic AI-generated website

How to make better websites with AI

AI still makes generic websites when you give it generic instructions. The difference now is that it can actually follow a strong design direction.

The quality of the result depends much more on the direction you give it. A specific visual style, described clearly and paired with one of the newest models, can produce something genuinely distinctive even if you aren't a designer.

That wasn't true six months ago.