dFusion AI

dFusion AI is building an open knowledge graph for the Agentic Web — a decentralized data layer designed for AI systems, enterprise workflows, and autonomous agents. The protocol addresses a fundamental gap: publicly scrapable internet data is running out, while 96% of valuable information lives in private chats and specialized sources that conventional AI training cannot reach.

The stack runs across three live layers. Vericore operates on Bittensor Subnet 70 as a semantic fact-checking layer — processing statements and returning evidence-based validation from real sources. The Social Truth DLP on Vana turns encrypted Telegram data into a community-owned dataset, rewarding contributors in $VFSN. Aethir provides decentralized GPU compute across the subnet infrastructure. EigenLayer-secured AI filters are under active development.

The platform currently reports 650,000+ contributors, 80M+ insights submitted, and a knowledge base described as 7x larger than Wikipedia with an 18% submission acceptance rate. The main protocol token, $DFSN, has not yet launched. At least 1% of the total $DFSN supply is reserved for Genesis contributors — those who submitted and had knowledge accepted during the testnet phase.

Co-founders: Roger Ying (CEO, Stanford MS&E, ex-PolicyDock) and Patrick De La Garza (CTO, ex-Zocdoc, Blockseer). Backed by Yuma — a Bittensor accelerator under Digital Currency Group (DCG).