SPINORIO
The AI workspace
for quantum computing
Work with an AI agent designed for quantum computing to develop code, run simulations, compile circuits with built-in AI-powered models and explore large-scale quantum optimization problems using the workspace’s solver. Keep your conversations, notebooks and results together in one workspace.
After the same RZZ angle normalization, Weylattice has lower circuit and two-qubit depth, with more gates. Both circuits reproduce the source state in an ideal simulation.
Circuit validation and downloads
Original Weylattice output followed by local RZZ angle normalization: X(a) · RZZ(−θ) · X(a) replaces each negative-angle RZZ(θ). All displayed RZZ angles satisfy 0 < θ ≤ π/2.
Gate names and couplings match the recorded IBM Fez target. The complete fixed experiment matches the source statevector after output-qubit remapping, to numerical precision. This is a simulation check, not measured hardware fidelity or an arbitrary-input equivalence proof.
All operations and active wires are retained. Only idle hardware wires are hidden. Metrics describe these normalized circuits; compile times describe the original runs.
Angles on the diagram are rounded to six decimals; exact stored values are available by selecting a gate and in the exports. Measurement labels c[i] identify output bits. Global phase: 1.5707963267949037 rad.
| Normalized circuit metric | Baseline | Weylattice |
|---|---|---|
| Circuit depth | 200 | 193 |
| Two-qubit depth | 94 | 87 |
| Two-qubit gates | 195 | 237 |
| Unitary gates | 655 | 807 |
| Original compile time | 0.062 s | 1.673 s |
About the recorded example
Recorded 16 November 2026. Same source circuit, IBM target snapshot and optimization level 2. Fixed QAOA parameters; no QPU execution or optimization of the MaxCut objective. Complete circuit diagrams include local RZZ angle normalization applied to both original outputs. Gate names, couplings and RZZ angles were checked; fixed-input statevector checks passed. Counts and depths include normalization; compiler timings do not. This does not establish arbitrary-input equivalence or hardware fidelity. A single example, not a general benchmark.
Download recorded dataBUILT AROUND YOUR RESEARCH
From a research question
to a working experiment
Work with your agent, edit the code yourself, and review each result in the same workspace.
Turn an idea into your next experiment
Describe what you want to explore. Work with your agent to develop code, use quantum tools and work through the next step, with the activity visible as it happens.
- Keep conversations and files organized in projects
- Inspect tool activity and review proposed file changes
- Choose your model and reasoning effort
I’ll create a notebook with three QAOA layers. Review the circuit construction before compiling for your target.
qc = QuantumCircuit(16, 16)
qc.h(range(16))
for gamma, beta in zip(gammas, betas):
for u, v, weight in edges:
qc.rzz(-gamma * weight, u, v)
qc.rx(2 * beta, range(16))
qc.measure(range(16), range(16))Take the code into your own hands
Open a notebook, edit a cell, or explore a variable in the Python console. Move between conversation and code as your experiment develops.
- Edit and run real Python notebooks in the cloud
- Use a console connected to your notebook kernel
- Work with Qiskit, PennyLane, Braket and pytket
qc = QuantumCircuit(16, 16)
qc.h(range(16))
for gamma, beta in zip(gammas, betas):
for u, v, weight in edges:
qc.rzz(-gamma * weight, u, v)
qc.rx(2 * beta, range(16))
qc.measure(range(16), range(16))QAOA structure and parameters
Weighted MaxCut · 16 qubits · 24 edges · p = 3. Initial H gates prepare |+⟩ on every qubit. Each layer applies RZZ(−γw) for every edge, followed by RX(2β) on every qubit.
C = Σ wᵢⱼ(I − ZᵢZⱼ)/2; U꜀ = exp(−iγC); Uᵦ = exp(−iβΣXᵢ).
γ = [0.31, 0.52, 0.73]; β = [0.19, 0.28, 0.37]. Fixed angles, without classical parameter optimization.
Angles on the diagram are rounded to six decimals; exact stored values are available by selecting a gate and in the exports. Measurement labels c[i] identify output bits. Global phase: 0 rad.
>>> qc.num_qubits, qc.depth()
(16, 36)Prepare circuits for your target
Use AI-powered quantum compilation within your workspace. Preserve the source circuit, select a target, and inspect the compiled output before taking the next step.
- Work with live IBM, saved or offline targets
- Inspect depth, gate counts and a comparable baseline
- Return completed circuits to your notebook
16 logical qubits · QAOA, 3 layers · 24 weighted edges
Circuit validation and downloads
Original Weylattice output followed by local RZZ angle normalization: X(a) · RZZ(−θ) · X(a) replaces each negative-angle RZZ(θ). All displayed RZZ angles satisfy 0 < θ ≤ π/2.
Gate names and couplings match the recorded IBM Fez target. The complete fixed experiment matches the source statevector after output-qubit remapping, to numerical precision. This is a simulation check, not measured hardware fidelity or an arbitrary-input equivalence proof.
All operations and active wires are retained. Only idle hardware wires are hidden. Metrics describe these normalized circuits; compile times describe the original runs.
Angles on the diagram are rounded to six decimals; exact stored values are available by selecting a gate and in the exports. Measurement labels c[i] identify output bits. Global phase: 1.5707963267949037 rad.
| Normalized circuit metric | Baseline | Weylattice |
|---|---|---|
| Circuit depth | 200 | 193 |
| Two-qubit depth | 94 | 87 |
| Two-qubit gates | 195 | 237 |
| Total gates¹ | 655 | 807 |
| Original compile time | 0.062 s | 1.673 s |
After RZZ angle normalization: lower circuit and two-qubit depth, with more gates.
About this compilation
- Recorded on 16 November 2026, using the real 156-qubit ibm_fez configuration fetched through Spinorio.
- Same source circuit, target snapshot and optimization level 2 for both original runs. Both outputs then received local RZZ angle normalization.
- Gate names, couplings and RZZ angle ranges checked. The complete fixed experiment matches the source statevector after output-qubit remapping, to numerical precision.
- Compilation only. No QPU execution, fidelity measurement or calibration-aware optimization.
- Fixed QAOA parameters; this example did not solve the optimization problem.
- ¹Total gates excludes measurements and barriers. Counts and depths describe normalized exports. Times describe the original compilation, excluding local normalization, queueing and API overhead.
- A single example, not a general performance benchmark.
Build a plan for your optimization problem
Bring optimization into the same research workspace. Prepare a model, review an execution plan, and inspect the results with Hilvarian.
- Define supported MaxCut, QUBO and Ising problems
- Review settings and resource estimates before a run
- Keep solver jobs and results with your project
Keep the experiment connected
Follow supported Qiskit activity from circuit states to processing records, jobs and available results. Keep the evidence and context together as your research develops.
- Inspect circuit history and linked execution records
- Organize research in projects with contributor branches
- Export records for review and reporting
- Circuit stateSource, structure and lineage
- Processing recordHow the circuit was transformed
- Jobs and available resultsExecution evidence in context

QUANTUMCLASS
Begin your next stage of quantum learning
Explore the latest research and developments in quantum computing through structured lessons, hands-on assignments, and practical implementation.
Explore QuantumClassA GLOBAL LEARNING COMMUNITY
Participants from universities
and organizations around the world



YOUR NEXT EXPERIMENT STARTS HERE
Bring your quantum ideas
into Spinorio
Open your workspace and start building with quantum tools.
Try Spinorio Find the right solution for your work