A distributed, open scraper specifically for LLMs would be interesting . Filtering for spam and useless data would be a challenge at scale, but would be cool