For our demo, we show how one capable computer can serve AI tasks to multiple users. A shared Mac runs real inference while other laptops submit documents and code through their browsers. Users can discover the machine, choose an available model, and follow their tasks through completion.
Our larger vision is a network where individuals, universities, organizations, and external compute providers can make spare computing capacity available to others. Stranded Compute would connect people who need AI compute with those who already have the hardware, expanding access beyond what someone’s own device can run.
The hackathon prototype establishes the foundation: worker discovery, model-aware scheduling, remote inference, and execution tracking. Growing this into a broader platform would involve working with infrastructure partners and early adopters to develop secure access, workload isolation, reliable service, and incentives for contributing compute
