Show HN Feels Increasingly Samey, and HN Is Arguing About the Shape of AI Design
Original: Scoring Show HN submissions for AI design patterns View original →
This Hacker News thread was lively because it took a fuzzy complaint and made it countable. Developers have been joking about vibe-coded landing pages for a while, but this post actually tried to score the look.
The linked write-up, Show HN submissions tripled and now mostly share the same vibe-coded look, examined 500 recent Show HN landing pages and checked them for 15 recurring patterns. The list is specific enough to be recognizable at a glance: Inter-heavy hero sections, purple accents, perma-dark pages with soft contrast, badge-above-heading layouts, icon-topped feature cards, gradient-heavy backgrounds, glassmorphism, and shadcn/ui fingerprints. Instead of screenshot judgment, the author used deterministic Playwright-based DOM and CSS checks, then manually QA’d the run for false positives.
The result gave HN something concrete to argue about. The post grouped sites into three buckets: 105 heavy slop pages with 5 or more patterns, 230 mild pages with 2 to 4, and 165 clean pages with 0 or 1. It also connected the visual sameness to a traffic shift on HN itself, noting that Show HN volume has surged and that moderators recently restricted Show HN submissions for new accounts. In other words, the piece was not just about taste; it was about what happens when faster building tools push more people toward the same defaults.
HN comments widened the argument in useful ways. Simon Willison pointed out that side projects are exactly where people will use AI assistance first, because time pressure is high and polish is secondary. Other commenters said the AI look is not the real problem by itself. The stronger tell is when the visual sameness signals shallow product thinking: lots of tidy cards, not much evidence that the author deeply considered the workflow underneath. A few also argued that every generation had its own template era and that today’s sameness may simply be the latest version of Bootstrap fatigue.
That mix of reactions is why the post resonated. It did not claim AI design is bad by definition. It claimed that default-heavy design choices are now easy to detect at scale, and that once everyone can ship faster, taste and specificity matter more, not less. HN seemed to agree on one thing: the web is filling up with pages that work well enough, and the next competitive edge may be the pages that stop looking like they were all prompted from the same starter pack.
Related Articles
The Reddit debate focused on whether an AI detector was being used as evidence or as an uncalibrated decision-maker.
HN focused less on the leaderboard and more on how refusals, tool loops, and account permissions shaped the result.
OpenAI’s June 3 blueprint turns state frontier-AI bills into a proposed federal template. The plan centers on CAISI, independent audits, severe-risk evaluations, incident reporting, model-weight security, and a broader government resilience strategy.