Tether has moved its peer-to-peer search project from concept to a working test. CEO Paolo Ardoino said the company’s P2P team stored the full Wikipedia archive across 100 distributed nodes and kept the data searchable through a simple interface. Hypersearch was disclosed in April.
Wikipedia Archive Runs on 100 Nodes
In the Sept. 8 test, Ardoino showed a search interface connected to 100 nodes. Wikipedia results appeared alongside network details, including which nodes took part. The setup is meant to scale horizontally to thousands of nodes. If some participants leave or fail, the system should stay available. Tether is using fault-tolerant structures to let data organize itself across the network. That design could reduce reliance on any single server and make search harder to censor.
The project uses HyperDHT, part of the Holepunch software stack. HyperDHT helps peers find each other and exchange information directly, without a central directory handling every request. Tether has not said when the search engine will launch publicly or how users outside its test environment will get access.
Recruiting for Distributed Search
Tether is hiring engineers for its P2P search team. The job listing asks for experience in distributed databases, conflict-free replicated data types, search architecture, and P2P networking. It also covers ranking algorithms, inverted indexes, lexical and semantic search, NAT traversal, encryption, and network protocol tuning. The listing names HyperDHT, Hyperswarm, Hypercore, Hyperbee, Hyperdrive, and UDX as technologies used by the team.
Search is not Tether’s first P2P software effort. Tether and Bitfinex backed Holepunch and its encrypted communication app Keet. Keet was built for direct audio calls, messaging, and file transfers over a distributed network. Tether later used similar ideas in Bitcoin mining. Its Mining OS came out as open-source software in February 2026 with encrypted P2P networking. A Mining Development Kit followed in April, offering hardware control and monitoring through JavaScript and React.
Questions Remain About Public Scale
Hypersearch applies the same distributed approach to information retrieval. Instead of keeping the searchable dataset and processing infrastructure in one place, it divides those functions among participating nodes. The Wikipedia deployment is a controlled test using a large document collection. Tether has not explained how the engine would crawl or index the wider web. It also has not detailed ranking at public scale or how it would handle spam, manipulated results, and conflicting data on an open network.
The search project sits alongside other Tether software outside stablecoins. Through QVAC, the company has worked on tools that run AI models locally instead of relying on centralized cloud infrastructure. In April, Tether released the QVAC SDK, an open-source toolkit for running AI applications on consumer devices. QVAC Fabric supports text, speech, vision, and translation workloads, with Holepunch tech used for P2P model distribution and delegated inference.
For now, the search engine remains under development. Ardoino’s latest demo shows Wikipedia running across 100 nodes, while Tether designs the architecture to eventually operate across thousands.
