Tether CEO launches centralized P2P search engine

TETHER tests decentralized point-to-point search technology: Wikipedia's full data successfully runs on a 100-node network

TETHER recently tested its decentralized point-to-point (P2P) search technology, storing complete Wikipedia archives on 100 distributed nodes. As the issuer of USDT, TEDA is developing a system that does not rely on centralized search infrastructure and is designed to maintain high availability.

Build an "unstoppable" decentralized search engine

According to Paolo Ardoino, CEO of TEDA, the company's P2P team is developing a product he calls the "Unstoppable Search Engine." The system is designed to scale to thousands of nodes and enables self-organization of data through a fault-tolerant structure.

Ardoino shared the latest test results on September 8, showing a work search interface connecting 100 nodes. The complete Wikipedia archive is distributed across the test network and remains searchable. Ardoino described the system's performance as "amazing." The demonstration was based on Hyperresearch, a decentralized search engine project that Ardoino first disclosed in April this year. The project uses distributed infrastructure rather than centralizing the search indexing and query process in the hands of a single centralized operator.

Fault-tolerant architecture and anti-censorship capabilities

In the latest test, TEDA distributed Wikipedia archives on 100 separate nodes and demonstrated the process of processing search requests over the network. The interface displayed by Ardoino not only displays Wikipedia search results, but also contains network information such as nodes participating in the test.

According to Ardoino, the system adopts a fault-tolerant architecture that allows data to be self-organized in a distributed system. This design is designed to increase the search service's resistance to censorship while reducing reliance on any single server or network participant. The architecture is designed to scale horizontally to thousands of nodes and maintain high availability even when individual participants leave the network or become unavailable.

Development work is still in progress, and Ardoino called the team's progress "impressive." However, TEDA has not announced a public release date for the search engine, nor has it detailed when external users will be able to access services outside the test environment.

Low-level support based on Holepunch technology

Hypersearch was first disclosed on April 7, when Ardoino described it as a decentralized search engine based on Distributed Hash Table (DHT) technology. DHT allows information to be stored and located among network participants without requiring a central directory to process all requests.

The project uses HyperDHT, which is part of the point-to-point technology associated with the Holepunch software stack. HyperDHT provides a way for peers to directly find and exchange information with each other, forming one of the infrastructure used by TEDA in its multiple software projects. Currently, TEDA is recruiting engineers for its P2P search team, with job requirements including experience in distributed databases, conflict-free replicated data types (CRDT), search architecture and P2P networking. Specific responsibilities cover areas such as ranking algorithms, inverted indexing, lexical and semantic search, NAT penetration, encryption and network protocol optimization.

Other technologies from the Holepunch family, such as HyperSwarm, HyperCore, HyperBee, HyperDrive and UDX, are also mentioned in the job list.

From communication to mining: TEDA's P2P layout extension

Search is not TEDA's first attempt to build software around a point-to-point architecture. Previously, TEDA and Bitfinex supported Holepunch and its encrypted communication application Keet. Keet is designed to support direct audio calls, messaging and file transfers over a distributed network without having to route traffic through traditional centralized infrastructure.

Later, TEDA applied similar concepts to its Bitcoin mining business. Its Mining OS was released as open source software in February 2026. It has encrypted point-to-point network capabilities and supports various operations from personal mining settings to industrial-grade facilities. In April of the same year, the company released the open source Mining Development Kit, which aims to provide hardware control and monitoring through JavaScript software development tools and React user interface libraries. The system is compatible with Windows, macOS and Linux, and is suitable for a variety of scenarios from home mining equipment to large mines.

Prior to this, TEDA's early Moria mining platform had used Holepunch technology to connect various components through a P2P model. Tests in 2023 showed that write operations require multi-signature approval, and the platform combines point-to-point communication with Internet of Things technology.

Beyond stablecoins: A new exploration of AI and distributed computing

The search project is one of a series of software products developed by TEDA outside of its USDT business. Its work has been extended into artificial intelligence, with the QVAC project developing tools aimed at running models locally to reduce reliance on centralized cloud infrastructure. In April this year, TEDA released the QVAC SDK, an open source toolkit for running AI applications directly on consumer-grade devices.

QVAC Fabric supports text, speech, visual and translation tasks, and leverages Holepunch technology for point-to-point model distribution and delegated reasoning. TEDA said its development plans include decentralized model training and fine-tuning, as well as toolkits for other computing applications. This architecture follows early efforts to move AI processing to user hardware. In March, TEDA announced that QVAC Fabric has fine-tuned models with up to 3.8 billion parameters on devices such as the Pixel 9, Galaxy S25 and iPhone 16, and has tested models with up to 13 billion parameters on the iPhone 16.

Currently, TEDA's search team is working on solving related distributed system issues, including copying data, networking and retrieving information. Its current engineering hiring specifically seeks developers familiar with Kademlia-style DHT systems, CRDT and Holepunch stacks, as well as traditional search concepts such as ranking, recall, accuracy and inverted indexing.

At this stage, the search engine is still in the development stage. Ardoino's latest demonstration shows that the complete Wikipedia archive is already running on 100 distributed nodes, while TEDA is designing the underlying architecture with the hope of eventually running on thousands of nodes.