Tether open‑sources QVAC SDK
Tether released QVAC SDK, an open‑source, cross‑platform AI framework designed to train and run models locally on devices rather than relying on centralized servers. The SDK targets on‑device workflows and could appeal to applications that need lower latency, privacy isolation, or offline capability. Making the framework open‑source increases the chance of community adoption and integration across edge devices. (x.com)
Most artificial intelligence still works like a vending machine in a warehouse: your phone sends a request to a distant server, the server does the work, and the answer comes back over the internet. On April 9, 2026, Tether said it wants that work to happen on the device in your hand instead. (tether.io) That shift is called on-device artificial intelligence, and it means the model runs on a laptop, phone, or embedded computer instead of a company’s cloud. Apple, Google, and Qualcomm have all pushed this direction because it cuts round-trip delay and keeps more data on the machine that created it. (siliconangle.com) A software development kit is the box of parts developers use to build an app without writing every screw and wire from scratch. Tether’s new box is called QVAC Software Development Kit, and the company released it as open source on GitHub this week. (github.com) QVAC says developers can run language models, speech tools, retrieval systems, and other artificial intelligence tasks across Linux, macOS, Windows, Android, and iOS with one stack. The public documentation describes it as a local-first system, which means the default assumption is “run here first,” not “send it away first.” (docs.qvac.tether.io) Tether first announced QuantumVerse Automatic Computer, the long form behind QVAC, on May 14, 2025 as a platform for artificial intelligence agents that live on user devices instead of big data centers. The April 2026 release turns that earlier pitch into a developer kit people can actually download and test. (tether.io) Under the hood, QVAC includes a component called QVAC Fabric, which outside reporting describes as a fork of llama.cpp. That matters because llama.cpp is one of the main open-source engines people already use to run large language models on ordinary hardware without renting cloud servers. (siliconangle.com) The toolkit also plugs in speech-to-text and translation engines, including whisper.cpp, Parakeet, and Bergamot, according to Tether’s launch materials and follow-up coverage. In plain terms, QVAC is trying to be one toolbox for text, voice, translation, and retrieval instead of a pile of separate parts taped together. (siliconangle.com) There is another piece that sounds more like Tether’s crypto background: peer-to-peer computing. QVAC’s GitHub page says a device can hand work to other peers for inference, which is closer to BitTorrent-style sharing than to the usual “everything goes through one company’s server” model. (github.com) That design opens a different set of uses than a giant cloud model. A factory tablet with no stable internet, a hospital device that should not send raw audio off-site, or a field tool that needs an answer in milliseconds all benefit when the model sits nearby instead of in a faraway data center. (qvac.tether.io) Open source does not guarantee adoption, but it changes who is allowed to tinker. By putting QVAC in public repositories with public docs, Tether is inviting outside developers to inspect the code, port it to more devices, and decide whether this becomes a real edge-computing stack or just another corporate experiment with a GitHub page. (github.com)
