We are exploring how quantum operations can become discoverable peer-to-peer services—routed across a global compute mesh, executed by specialized nodes, and assembled into complete scientific results.
Quantum program
↓
Split into executable circuit fragments
↓
Discover compatible peers through libp2p
↓
Route fragments to distributed quantum services
↓
Execute on simulators today—and quantum processors tomorrow
↓
Reassemble, verify, analyze, and share results through IPFS
Nodes advertise capabilities through GossipSub, receive circuit fragments over libp2p streams, and return partial quantum states to a coordinator. Qiskit currently provides execution and analysis while IPFS enables content-addressed sharing of circuits and research artifacts.
- Financial modelling — QAOA portfolio optimization, option pricing, risk analysis, and quantum-versus-classical comparisons.
- Drug discovery — molecular simulation, quantum chemistry workflows, candidate scoring, and distributed scientific search.
- Scientific computing — reusable quantum operations exposed as network services for researchers and autonomous agents.
- Open quantum infrastructure — peer discovery, execution planning, fault-aware routing, provenance, and independently operated compute nodes.
Current Qiskit statevector experiments compare QAOA portfolio optimization with Simulated Annealing:
| Portfolio size | Classical (SA) | Quantum (QAOA) | Result |
|---|---|---|---|
| 10 assets | 20 ms | 1,500 ms | Classical 75× faster |
| 20 assets | 600 ms | 1,700 ms | Classical 2.8× faster |
| 40 assets | 6,000 ms | 1,900 ms | Quantum 3.2× faster |
| 60 assets | 20,000 ms | 2,100 ms | Quantum 9.5× faster |
One important finding: roughly 97% of current QAOA runtime is classical parameter search, not circuit execution. That means distributed nodes alone do not create quantum advantage—the opportunity comes from better optimization methods and more favorable scaling as problems grow.
These are simulator-based research benchmarks, not claims of production quantum-hardware advantage. Results depend on the workload, solver, configuration, and machine.
