Roox aims to make AI inference verifiable, not just executable.
A future receipt can bind the request, response hashes, model, provider, measured usage, quoted price, timestamps, nonce, and worker signature into one record that can be checked before payment.
Choosing an AI provider should not mean accepting whatever infrastructure is available.
Roox is designed around marketplace choice, exposing model, GPU, region, price, latency, uptime, and worker availability so developers can route requests around their own constraints.
Coming to @RobinhoodApp
Roox separates four parts of AI inference:
Discovery, execution, verification, and payment.
A request is matched to a GPU worker, the worker returns output with a signed receipt, and a verifier can authorize settlement only after the completed call checks out.
Introducing Roox
AI inference is becoming a commodity, but access to compute is still fragmented.
Roox proposes an open marketplace where developers can choose providers by model, GPU, price, latency, hardware, and region, while independent GPU owners can earn from idle capacity.
Launching soon on @RobinhoodApp…