Needle: The benchmark your search engine can't memorize

Keenable, a search startup that raised $26M from Accel and Conviction, introduced NEEDLE, an open-source benchmark designed to evaluate search engines for AI agent use cases. Unlike static benchmarks that are prone to overfitting, NEEDLE is a live benchmark drawing from real agent search logs to measure how well search engines handle fresh and niche information that AI agents need to complete real-world tasks. The benchmark runs continuously and publicly, allowing anyone to reproduce results and contribute improvements.