your browser tab reads crypto docs to build an open AI dataset.

yes. the crawler is plain javascript running in a sandboxed web worker in this tab. it can only ask our server for pages on a fixed list of public crypto sites (docs, EIPs, github readmes, dao forums). the server fetches them for you with size and time limits and blocks anything private. nothing gets installed and your files, wallet and other tabs are never touched.

a crawler opens a web page, pulls out the readable text and follows the links on it to find more pages. search engines crawl the whole web. hivemind crawls a fixed list of public crypto sites, and your tab does the work while it's open.

your crawler is the small program running in this tab when you press start. it's shown as a unique spider (built from its name) that walks around the page and carries each page it reads to the queen. rarity is cosmetic only.

a community AI model that will be trained on the hive's dataset. training hasn't started yet, so the page says "training pending" until a real run begins.

the dataset is every cleaned, de-duplicated page the hive has read, free to download. tokens are the small chunks of text AI models read (roughly 4 characters each), so dataset size is counted in tokens.

each UTC hour is a round. crawlers with 25+ new pages that pass the server's spot-check in that hour qualify, and that hour's creator fees are split by verified pages × boost. see the rounds panel.

a boost multiplies your round share based on your live $hive balance: 1x under 1M, 1.5x at 1M+, 2x at 5M+. "verify wallet" means signing a short message in your wallet to prove you own the address you crawl with. it's free and never a transaction.

optional. holding it boosts your share of each round and lets you vote on which sites get read next. crawling never needs it. ca coming soon, so boosts and pots show "pending". not financial advice.

about every 2 seconds it asks for a page, gets the page's text through our proxy, strips the html, fingerprints it (sha-256) so duplicates get skipped, counts roughly how many tokens it holds, and sends that text back. that's all it does: some network requests and a bit of cpu.

no money, no wallet, no signup. it uses a little bandwidth (pages are usually tens of kb, capped at 1.5 mb) and a little cpu while the tab is open. on a metered mobile connection you might want to stop it after a while.

yes. press stop or just close the tab. nothing keeps running afterwards and there's nothing to uninstall.

pages are deduplicated and stored as an open dataset that anyone can download as jsonl. it's meant for training the queen. no training run has started yet.

no. crawling is free and always will be. holding $hive boosts your hourly share and lets you vote on what gets read next. both go live when the token launches (ca coming soon).

hivemind is a sister hive to crawlnet, and was inspired by it: they pretrain, we crowd-crawl. we're an independent community project, not affiliated with or endorsed by crawlnet, and we're cheering them on.

paste your solana address as your identity, press start, then press "verify wallet" and sign the message in phantom (it's free, not a transaction). every UTC hour, crawlers with 25+ new verified pages split that hour's pump.fun creator fees by verified pages × holder boost (1x under 1M $hive, 1.5x at 1M+, 2x at 5M+). see the rounds panel and /api/rounds.

no. the pot is only what creator fees actually come in that hour. it can be tiny or zero, and the ca isn't live yet, so pots show "pending". payouts are sent by hand from the public payout plan, not automatically. none of this is financial advice.