Scoring Show HN submissions for AI design patterns

Article

TL;DR

67% of recent Show HN sites show 2+ AI design patterns; heavy slop in 21% of submissions.

Key Takeaways

• 21% of Show HN sites score 5+ AI patterns (heavy slop); only 33% score clean
• Patterns like Space Grotesk, left-border accents, and icon-card grids predate AI — attribution is noisy
• Signal-to-noise collapse on Show HN: AI lowers proof-of-work, making real commitment harder to detect

Discussion

Top comments:

• [jerf]: In 2026, 10k lines proves nothing — proof-of-work collapsed with AI coding tools

  In 2026, 10,000 lines of code means they spent a minimum amount of money and almost no time.

• [dang]: HN already restricts new accounts via showlim — that explains the recent downtick
• [nomdep]: Most listed patterns were common before AI — blame TailwindUI and shadcn, not LLMs
• [tptacek]: AI look-and-feel is fine; the real problem is vibe-coded substance with no commitment

Discuss on HN