$CRAWL 2.6m dyor BXoHJddsWJLHtAopeiSbKUSELsu8hSFMs8baGMDkpump okx wallet: Use a swarm of crawlers to scrape crypto and Web3 websites, then use the collected content to train an open-source large language model from scratch that has only been exposed to crypto-related content. CRAWL is the associated token, used to pay for generating crawlers and to distribute a share of the fees. Here’s how the mechanism works: The crawlers do the work. Each crawler is a headless browser powered by a DeepSeek agent that navigates real websites, reading whitepapers, documentation, governance pages, code, and other content. It keeps only pages relevant to crypto and Web3 and discards everything unrelated. The data is then sent to “Queen.” All the content collected by the crawlers is cleaned, deduplicated, and tokenized before being aggregated into Queen’s training dataset. Queen is the model they intend to train. Users who want to participate can burn approximately 1 million CRAWL tokens to generate a crawler registered under their name. Twenty percent of the creator fees are distributed equally among crawler owners every 24 hours, with payments made in SOL. The remaining 80% is used to cover crawling costs and GPU training. The website allows users to watch in real time which page each crawler is reading. Once training is complete, the model weights are planned to be published on Hugging Face, with contributors credited. In short, CRAWLNET can be understood as an experiment in “publicly crawling crypto websites → accumulating training data → training a specialized small model,” combined with a token mechanism in which users burn tokens to obtain a share of crawler ownership and receive a portion of the fees.