Scoring Show HN Submissions for AI Design Patterns
Adrian Krebs developed a method to automatically detect AI-generated design patterns in Show HN submissions on Hacker News. He scored 500 landing pages against 15 common AI design patterns, such as colored left borders, Inter font usage, gradient backgrounds, and icon-topped feature cards. The analysis was triggered by a perceived increase in generic, sterile designs following the release of Claude Code. Krebs used Playwright for headless browser testing and deterministic CSS/DOM checks to avoid LLM-based judgment. Results were grouped into tiers based on pattern count. The project was human-written with AI-assisted scoring and analysis.
Key facts
- Adrian Krebs scored 500 Show HN landing pages for AI design patterns.
- 15 common AI design patterns were identified, including colored left borders, Inter font, and gradient backgrounds.
- Claude Code led to a large increase in Show HN submissions.
- HN moderators restricted Show HN submissions for new accounts due to the increase.
- The scoring method used Playwright for headless browser loading and deterministic CSS/DOM checks.
- False positive rate was estimated at 5-10% based on manual QA.
- Patterns were grouped into three tiers based on how many of the 15 patterns were triggered.
- The post is human-written, but the scoring and analysis were AI-assisted.
Entities
Artists
- Adrian Krebs
Institutions
- Hacker News
- Show HN
- Claude Code