Hacker News
I Scraped 1.94M Airbnb Photos — CLIP + Claude Haiku Vision at Scale on Burla's Distributed Platform
A developer scraped 1.94M Airbnb listing photos across 119 cities and 4 quarterly snapshots, then scored them with CLIP and had Claude Haiku Vision double-check suspicious results — all parallelized on Burla's distributed compute platform (~1.7K CPU workers for download/scoring, 20 A100 GPUs for embeddings). The project also scored every review and reranked the weirdest 12K with Haiku. A practical demonstration of vision AI at dataset scale. 68 points on HN.
↳ Follow the thread