Physa explores how idle devices can cooperate to form a shared compute field.
Computers and phones, already today. Work travels from neighbor to neighbor like a wave, and comes back as one assembled result.
The requester states an intent and its constraints. It does not dispatch tasks machine by machine. The field carries the work, solves it, and sends back the assembled answer.
O releases a wave carrying the work, its constraints and the rules for combining results.
Each node reads the wave, keeps what it can handle, and relays the rest to its neighbors.
Fragments run locally, in parallel, inside isolated runtimes.
Partial results meet inside the field. Intermediate traffic never reaches the requester.
O receives one final result, checked against the expected answer.
No central scheduler. Every node applies the same local rule, and global resolution emerges from local decisions.
A graph and a compiler infer how a job can be split safely. Anything with an unknown structure stays whole.
A compact kernel owns the live state of each wave. Transport and storage sit behind tested boundaries.
Rust workers run ONNX models and WebAssembly modules locally, isolated from the host device.
Every run is checked against the same computation on a single machine. Correctness, duplicates and coordination cost are measured separately.
Affine ONNX models, a transformer encoder and the WebAssembly VaR were all recognized by structure inference, with every result matching the reference.
The wave logic lives in one module with a boundary enforced by tests. Transport only consumes what the kernel emits.
Physa targets divisible work where computation outweighs communication. We publish the limits as openly as the wins.
Simulation, optimization, scoring and batch AI workloads. The goal is many useful computations running side by side.
A small language model split into six slices: every hop adds latency. Sequential inference is not where a field shines.
Per token, best run, SmolLM2 on a shared cloud box.
Physa is research in progress. The engine exists. The network is what we are building next.
Protecting both the device that contributes and the work it receives. Encrypted transport, signed waves, real isolation, participant admission. The threat model comes first, mechanisms second.
Computers and phones already join the field. Staying reliable at scale through mobility, battery limits and dropped connections is the open question.
Current evidence is bounded. Thousands of nodes, varied workloads and a general economic advantage are still to be established.
Separate the specification from the implementation, so others can build compatible nodes.
Define the threats to contributors and requesters, then choose the mechanisms.
Cut fast path costs, test on separate machines, then widen devices and network conditions.
A useful computation, a reproducible reference, success criteria set before the run.