The QVAC Blog
The complete guide to QVAC's 15 AI capabilities
Every AI capability in the QVAC SDK — text generation, embeddings, RAG, fine-tuning, multimodal, image and video generation, transcription, TTS, voice assistant, translation, BCI, VLA, OCR, and image classification. All on-device.
How QVAC actually works under the hood
A technical walkthrough of QVAC's internals — the Bare worker, the singleton RPC client, lazy initialization, and the four-phase lifecycle from loadModel() to close().
QVAC can run a brain-computer interface
QVAC ships brain–computer interface transcription — decoding multi-channel neural signals into text, entirely on-device. Here's what that is, why local processing matters more here than anywhere else, and where it points.
QVAC for robotics: vision-language-action explained
QVAC ships vision-language-action support — turning camera frames, robot state, and natural-language instructions into robot control, entirely on-device. Here's what VLA is and why local inference is non-negotiable for robots.
QVAC explained simply: a beginner's guide
QVAC is Tether's open-source engine for running AI directly on your device — no cloud, no account, nothing leaving your phone. Here's what it is and why it matters, in plain English.
Delegated inference: how your phone borrows your laptop's GPU
QVAC lets a device delegate AI inference to a peer over an encrypted P2P connection — your phone offloading to your desktop, no cloud involved. Here's how delegated inference and blind relays work.
On-device RAG over your own notes
A technical look at how localhost/AI is building retrieval-augmented chat over private notes using QVAC's on-device embeddings, vector search, and LLM — with nothing leaving the device.
Building on-device voice assistants with QVAC
QVAC chains transcription, text generation, and text-to-speech into a real-time voice loop that runs entirely on your device. Here's how to build a voice assistant that never sends your voice to the cloud.
Private finance without bank linking
Most finance apps require linking your bank through a third party like Plaid. Local Money (coming soon) won't — here's why on-device finance without a live bank connection is more private, and how it will work.
Is QVAC production-ready? An honest assessment
We're building consumer apps on QVAC. Here's a candid evaluation of the SDK's maturity — the repo signals, what works, what doesn't, and who should and shouldn't build on it today.
Inside the QVAC monorepo
A complete guide to QVAC's repository structure — every package in /packages, what it does, and how the SDK, addons, core libraries, and tools fit together.
Why on-device AI? QVAC vs cloud AI, answered simply
Quick, plain answers to the most common questions about QVAC vs cloud AI — is it more private, does it work offline, is it cheaper, and what's the catch.
QVAC vs private AI assistants
Many assistants describe themselves as private. Very few run on your device. Here is how to tell the difference.
QVAC developer resources
Where QVAC's documentation and source live, what to read in which order, and the one thing to build before anything else.
QVAC vs local LLMs
Running a local LLM gives you text generation. A real application also needs search, document reading, and speech. QVAC provides all of it on-device through one interface.
Best QVAC apps
The applications built on QVAC, from Tether's own Workbench and Health to independent local-first tools. What each does, and how to tell a genuine local app from a marketing claim.
QVAC vs Android AI
Android's on-device AI is Google's model, on Google's terms. QVAC is open source, model-agnostic, and assumes no account. A comparison of the two approaches.
QVAC vs Apple Intelligence
Apple runs some AI on your iPhone and sends the harder requests to its own servers. QVAC sends nothing anywhere. The difference is where the boundary sits, and who controls it.
QVAC Health guide
QVAC Health aggregates wearable data into an encrypted offline dashboard, with on-device AI for natural-language logging and meal photo analysis. Here is how it works and why its architecture matters.
QVAC vs LM Studio
LM Studio is a desktop application for downloading and chatting with local models. QVAC is a platform for building applications on them. Here is the practical difference.
QVAC Workbench guide
QVAC Workbench is Tether's local AI app for phones and desktops. What it does, what Delegated Inference means, and how to use it to choose a model.
QVAC vs Ollama
Ollama runs open models on your desktop. QVAC is an SDK for shipping local-first apps, including on phones. They solve different problems, and both can be the right answer.
QVAC local AI use cases
Practical applications where on-device AI outperforms the cloud — and the cases where it honestly does not.
QVAC vs OpenAI
How QVAC's on-device platform compares to OpenAI's hosted models — and the compatibility layer that lets you use both.
QVAC agents explained
An agent takes actions, not just generates text. Running one locally changes its risk profile entirely, because a local agent reads your files where they already are.
QVAC vs cloud AI
The difference between running a model on your device and sending your data to someone else's. A structural comparison, not a feature list.
How to build a QVAC app
A practical sequence for building your first local-first AI application on the QVAC SDK, and the constraints to plan around.
QVAC SDK explained
The QVAC SDK is the main entry point for building local-first AI apps. What it exposes, how models are loaded, and what to expect when building on it.
What is QVAC?
QVAC is Tether's open-source platform for running AI models directly on your own device, with no cloud and no account. Here's what it is and why it matters.