Quantum architecture program. What we tested

A focused IBM Heron r2 program (June 2026). Quantum Polycontextural Computing (QPC) is a gate-level architecture: many logics co-resident in one circuit, coupled, measured once. The goal is demonstration and falsification on named jobs: ICC, Joint Structure, Cerrado, MCGS, E8.

Plain English

What this program is
The hub for QPC’s IBM Heron r2 architecture tests (June–July 2026): separability, Joint Structure Challenge, Cerrado, MCGS, E8.
Why it exists
To show true parallel polycontextural execution on gate hardware: one circuit, co-resident logics, public job IDs.
What QPC computed
Coupled vs separable vs intracontext arms · merge-failure checks · real Goiás portfolio data · gauge-sector extension. IBM job IDs on each report.
Headline result
Architecture signature on ibm_fez · JSC L1–L5 pass · MCGS 3/3 · separability toy 3/3 · coupling ablation on two Heron r2 chips.
What we report
Falsifiable architecture evidence on raw counts: ICC gaps, merge-failure, named rankings. Metric and baseline stated on each page.

Flagship: Joint Structure Challenge. Six-level pre-registered ladder (ICC + merge failure). L1–L5 pass on IBM Heron r2 · ibm_fez. L6 is the open vendor dare.

What QPC means here (one sentence)

Multiple objective contextures execute in parallel inside one quantum circuit, linked by transjunction gates while alternatives stay open. Then you measure once on the gate QPU.

Carbon, biodiversity, and social impact sit in that one IBM job. The ranking comes from the coupled state.

Why an external AI review mattered

An independent review first treated stacked Hamiltonian layers as QPC. The lock is: parallel contextures in one circuit, coupled by transjunctions. After the true-parallel jobs and IBM IDs, the same review proposed a separability test: measure correlation that independent jobs miss.

We ran that test on IBM Heron r2 · ibm_fez and published the outcomes on the separability and JSC pages.

Workstreams. What each one proves

WorkstreamQuestionStatusWhat it demonstrates
Separability (ICC) Does one coupled job produce joint structure that three independent jobs cannot? Demonstrated One coupled job vs independent jobs. Toy 3/3 seeds; Cerrado n=12; Heron n=28 bio↔social boundary 0.098.
Cerrado portfolio Can real Goiás carbon-credit data run as one polycontextural workflow? Demonstrated Same published metric, open data, auditable IBM jobs. One submission. Architecture stands on the ICC pages; KPI on the Cerrado compare page.
E8 lattice test Can eight parallel certificate checks run as one coupled Heron r2 job? Demonstrated 152Q, one job d8o47bjqv2lc7389jo60. E8 is the known answer key. Plain-language report on site.

Results in plain language

1. Separability. “Is it really one quantum thing?”

We run two arms on the same IBM chip:

We compute inter-context correlation (ICC): do decisions on one objective line up with the next more than independent runs would?

TestICC gap (coupled − separable)Verdict
Toy instance (seeds 7, 11, 42)0.052 – 0.106Strong pass 3/3
Cerrado real data n=120.088Pass
Cerrado Heron n=28 (mean)0.031Below mean threshold
Cerrado Heron n=28 (bio↔social boundary alone)0.098Heron-scale pass

Value: This is the architecture signature. The hardware returned coupled quantum structure on named jobs.

2. Cerrado. “Can a real business task use one submission?”

Brazilian Cerrado: pick towns that jointly score carbon, biodiversity, and social impact (Ribeiro 2026, open GitHub data). Same portfolio formula as the paper.

Value: Gate-native multi-objective execution with job IDs on file. Architecture on raw ICC; portfolio KPI on the compare page.

3. MCGS. Gauge-sector motivated ICC

Three contextures with even / odd / bridge weight priors, motivated by the DFL lattice-gauge paper. IBM Heron r2 · ibm_fez: raw ICC gaps 0.349 · 0.243 · 0.131, 3/3. See MCGS.

Value: The same ICC stack on a new instance, auditable job IDs.

4. E8 lattice test. Multi-check search on Heron r2

In 8D, the densest sphere packing (E8) was proved by Viazovska using several coordinated checks. QPC runs that computational pattern on IBM: eight parallel certificate contextures in one 152-qubit job. E8 is the known answer key.

Value: QPC’s multi-context architecture on a structured hard problem. Same idea as Cerrado’s three objectives, different domain. Plain-language explainer →

How computing works (shared pipeline)

1. Load published or synthetic instance (Cerrado data, toy weights, antenna graph) 2. Compile K parallel context blocks into one gate-native workflow (proprietary schedule) 3. Run coupled (one IBM job) vs separable control (matched multi-job baseline) 4. Primary metric: inter-context correlation (ICC) gap on hardware counts 5. Secondary metrics: portfolio / placement scores with standard decoders (separate claims)

Compilation details, bridge wiring, and readout transforms stay with QPC. Contact readytogo@quantumpolycontextural.ai under NDA.

For partners. Resilience and two lanes

QPC runs a resilience stack after IBM jobs (readout correction, shot aggregation, optional extrapolation). Architecture evidence (ICC, Joint Structure Challenge, MCGS) stays on raw hardware counts. Customer-facing scores live on industry pilots with declared decoders. That is the two-lane split.

The add-on is documented here: QPC Noise Reduction Add-on. Partners under NDA: readytogo@quantumpolycontextural.ai.

Reports & upload pack

Separability. Full results and methodology Cerrado. Optimization architecture compare Real-world replication hub Comparable benchmarks index Special Report: QPC vs D-Wave E8 lattice extremal report Hostinger upload list (June 2026)