Crawlnet is a crypto-native project building Queen, a large language model pretrained from scratch on information gathered from the crypto web. Its headless crawlers visit public-facing resources such as whitepapers, documentation, governance materials and code using real browsers. The pages they collect are cleaned, deduplicated and tokenized to form Queen’s training dataset. The project says creator fees fund the GPUs and that each funded run retrains the model from zero, with the aim of making it larger over time. Queen’s model weights are intended to be released publicly.

The Crawlnet website presents a live dashboard for monitoring crawler activity, pages read, dataset tokens, domains and compute funding. It also describes a system for spawning sub-agents, or crawlers, to contribute data. A $CRAWL token is associated with the project, and the site states that owners split 40% of fees every 12 hours. Together, Crawlnet combines web crawling, model training and crypto-based funding in an open-data approach to developing an LLM for the crypto ecosystem.