December 29, 2025
Shelf Esteem Wars
Show HN: Vibe coding a bookshelf with Claude Code
Book hoarder tames 500 titles with AI; fans cheer, skeptics ask if taste still matters
TLDR: A coder used Claude Code to catalog roughly 500 books from photos, with a few manual fixes for accuracy. Commenters loved DIY “user-of-1” power but clashed over scale limits and whether human taste will stay valuable as AI handles execution, with some predicting taste won’t matter for long.
A book lover snapped 470 pics of their messy collection and let Claude Code do the dirty work—naming authors, titles, and covers with about 90% accuracy, then fixing the rest by hand. The community didn’t just clap; they turned it into a mini culture war. The feel-good camp swooned over “user-of-1” personal software: apps that do exactly what you want, running locally, no corporate bloat. One commenter called this the sweet spot for vibe coding—small projects where AI handles the busywork while humans steer the ship.
Then came the spicy takes. A linked think piece on “intent vs execution” kicked off big-brain energy, while a realist warned that once projects get complex, AI can spiral into subtle bugs. The hottest debate: does human taste still matter? One voice argued taste is our last advantage; another shot back that it won’t last for long as AI learns the vibe too. Meanwhile, everyone cracked up at the model confidently mislabeling a novel as a 1987 Soviet agricultural manual—instant meme fuel. The final verdict? People love the aesthetic of AI doing the grunt work, but they’re split on whether the human remains the creative director or just the person pressing “run.”
Key Points
- •The author cataloged a personal library using 470 book photos and AI-driven scripts built with Claude Code.
- •OpenAI’s vision API extracted author, title, and publisher data, achieving about 90% accuracy; remaining errors were corrected manually.
- •A second script processes new books through the same pipeline for ongoing updates.
- •Covers were fetched via Open Library’s API, then improved with a quality-scoring pass and fallback to Google Images via SerpAPI; about ten covers were manually edited in Photoshop.
- •The final UI began as a grid of covers, demonstrating a practical personal app where AI handles execution and the user applies judgment for edge cases.