Show HN submissions tripled and now mostly have the same vibe-coded look

Source: Hacker News

When browsing Hacker News, I noticed that many Show HN projects now have a generic, sterile feeling that suggests they are purely AI‑generated. I wondered if we could automatically quantify this subjective feeling by scoring 500 Show HN pages for AI design patterns.

Claude Code has led to a large increase in Show HN projects—so much so that the moderators of HN had to restrict Show HN submissions for new accounts.

Here is how Show HN submissions increased over the last few years:

That provides plenty of pages to score for AI design patterns.

AI design patterns

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. After asking other designers, the common AI patterns can be grouped into fonts, colors, layout quirks, and CSS patterns.

Fonts

• Inter used for everything, especially the centered hero headlines
• LLMs tend to use certain font combos like Space Grotesk, Instrument Serif, and Geist
• Serif italic for one accent word in an otherwise‑Inter hero

Colors

• “VibeCode Purple”
• Permanent dark mode with medium‑grey body text and all‑caps section labels
• Barely passing body‑text contrast in dark themes
• Gradient everywhere
• Large colored glows and colored box‑shadows

Layout quirks

• Centered hero set in a generic sans
• Badge right above the hero H1
• Colored borders on cards, on the top or left edge
• Identical feature cards, each with an icon on top
• Numbered “1, 2, 3” step sequences
• Stat banner rows
• Sidebar or nav with emoji icons
• All‑caps headings and section labels

CSS patterns

• shadcn/ui
• Glassmorphism

Example screenshots

Badge above the Inter hero.

Same, different page.

Colored border on top.

Icon‑topped feature card grid.

Gradient background + glassmorphism cards.

Detecting AI design in Show HN submissions

To systematically score these patterns, I processed 500 of the latest Show HN submissions:

• A headless browser (Playwright) loads each site.
• An in‑page script analyzes the DOM and reads computed styles.
• Every pattern is a deterministic CSS or DOM check; no screenshots are taken, and no LLM judges the visuals.

This method inevitably yields some false positives; manual QA suggests an error rate of about 5‑10 %. If there is interest in open‑sourcing the scoring code to replicate or improve the run, let me know.

Results

A single pattern doesn’t necessarily indicate AI‑generated design, so sites were grouped into three tiers based on how many of the 15 patterns they trigger:

• Heavy slop (5 + patterns) – 105 sites – 21 %
• Mild (2–4 patterns) – 230 sites – 46 %
• Clean (0–1 pattern) – 165 sites – 33 %

Is this bad? Not really—just uninspired. Validating a business idea has never been about fancy design, and before the AI era everything looked like Bootstrap. There is a difference between crafting a custom design and shipping with whatever defaults an LLM outputs, just as there was pre‑LLM when using CSS/HTML templates.

I expect designers will eventually return to creating beautiful, distinctive designs to stand out from the slop. On the other hand, it remains unclear how much design will matter once AI agents become the primary users of the web.

This post is human‑written; the scoring and analysis were AI‑assisted.