Pre-commercial computing platform
Benchmarked in joules per solution, not qubit counts.
QWAV explores p-adic ultrametric architectures as an alternative to the qubit-gate-circuit model. The approach draws on Ostrowski's theorem for intrinsic error protection and targets room-temperature operation. Every performance claim is benchmarked with JPCUB, an open, falsifiable measure of energy per correct answer.
Try it
Ten interactive demos
Each demo runs in your browser and illustrates one idea from the papers. They are teaching tools, not hardware measurements.
Staircase redundancy vs Archimedean codes
Bruhat-Tits quantum error correction, side by side with the standard code families.
Not checked by an independent implementation. Record
Virtual qubit showdown
An ultrametric tree against the surface code on the same noisy task.
Not checked by an independent implementation. Record
Error confinement
Watch Bruhat-Tits error suppression keep a fault from spreading.
Verified: 11 checks, breadth-first id numbering and level counts; explicit tree expansion vs closed form. Does not certify any quantum state, hardware result or advantage claim. Record
The tree and its shadow
An ultrametric phase diagram you can move through.
Not checked by an independent implementation. Record
Ultrametric convergence
Variance collapse under coarse-graining, step by step.
Verified: 6 checks, seeded generator re-implemented, direct block means, law of total variance. Does not certify any quantum state, hardware result or advantage claim. Record
Tree distance sandbox
Cophenetic, Euclidean and ultrametric distances on one tree.
Verified: 7 checks, base-p digit strings (prefix = LCA depth). Does not certify any quantum state, hardware result or advantage claim. Record
Hardware pathway visualizer
A 40-atom neutral-atom tree lattice, laid out to explore.
Verified: 8 checks, explicit tree expansion vs closed form; level counts of a 3-ary tree of depth 3. Does not certify any quantum state, hardware result or advantage claim. Record
Q-PNA classifier playground
Glass-box classification: see why each label is chosen.
Verified: 6 checks, Lance-Williams and Kruskal clustering, two-pass Pearson. Does not certify any quantum state, hardware result or advantage claim. Record
Monna map perspective
A renderer for the Monna map, from the UMP.010 C5 paper.
Verified: 5 checks, exact rational arithmetic (fractions) on p-adic digits. Does not certify any quantum state, hardware result or advantage claim. Record
Ultrametric tree universality
A cross-domain explorer of the same tree structure.
Not checked by an independent implementation. Record
The benchmark
Energy per correct answer
JPCUB measures total system energy per correct answer: memory, input and output, cooling and power conversion included, not only the processor. Do not trust anyone's numbers, ours included: measure any platform yourself in joules per correct answer.
JPCUB P0 protocol< 10−3 J
Landscape
How platforms compare on energy
QWAV does not compete on qubit counts. Only IBM carries a published P0-protocol measurement; the others are conservative system-level upper bounds from the competitive landscape paper (JPCUB Competitive Landscape).
| Platform | Approach | Operating temperature | Error correction | JPCUB | Basis |
|---|---|---|---|---|---|
| IBM | Superconducting transmon | ~15 mK | Active (surface code) | 0.89 J per solution | published (JPCUB P0) |
| Superconducting | ~15 mK | Active (surface code) | ~0.05 J per solution | estimate (landscape v2.3) | |
| Rigetti | Superconducting | ~15 mK | Active | ~0.61 J per solution | estimate (landscape v2.3) |
| IonQ | Trapped ions | Room temperature | Active | ~16.3 J per solution | estimate (landscape v2.3) |
| D-Wave | Quantum annealing | ~15 mK | None (annealing) | ~50 to 200 J per optimisation | estimate (landscape v2.3) |
| QWAV | p-adic ultrametric (qudits) | Room-temperature target | Intrinsic (Ostrowski) | < 10⁻³ J per solution | design target, not measured |
Architecture
The QWAV stack
From p-adic mathematics to applications. These layers are published as research papers; the hardware remains pre-commercial.
| Application | QWAV SDK (Python), QWAV Cloud API (REST), JPCUB benchmark dashboard |
|---|---|
| Compilation | Fontaine-stack compiler: problem, p-adic encoding, Bruhat–Tits building, ZBW observable, readout |
| Runtime | Adelic QEC (Ostrowski intrinsic protection), ZBW engine, Bruhat–Tits readout protocol |
| Physical | Trapped-ion Dirac simulator, room-temperature adelic nuclear-spin qubit (v2.0), 343-qubit tree topology |
Research
Built on published research
The work behind QWAV is by Rowan Brad Quni-Gudzinas and published by QNFO, an independent research imprint. Every paper is open and can be checked independently.
JPCUB P0: the metric
Definition, measurement protocol and anti-gaming provisions. Worked example: IBM Eagle at 0.89 J per solution.
The physics of computation
Landauer, Margolus–Levitin and Bremermann limits; joules per solution as the falsifiable criterion for physical computational advantage.
JPCUB competitive landscape
System-level estimates for 17 platforms from published specifications and third-party data.
Problem-substrate mapping
A framework for matching computational problems to physical substrates before investing in them.