Turn an algorithm into a reviewable experiment
Choose a bounded preset, seed, shot count, noise model, and resource contract; then compare three execution lanes produced by the real local N/M runtime.
Ideal, noisy, and mitigated experiment
Experiment Studio 0.1 creates a stable plan identity before running the same algorithm through bounded local simulator lanes.
All values are finite-shot local simulator evidence. This is not a QPU run, device calibration, hardware benchmark, or fault-tolerance claim.
Noise experiment series
Run the same workload at 2–5 evenly spaced noise probabilities. The model, seed, shots and per-run budget stay fixed. Runs execute sequentially in your browser.
Ready to configure a series.
Saved series
Save completed rows together in this browser (up to 10 series). Source code is excluded. Browser data removal also removes these records. Opening restores the table; inspecting a saved row reruns it. Projects require the matching local workspace source.
Export one saved series as JSON, including partial results. Import validates the file before saving; it does not run experiments or overwrite different evidence with the same ID. Project source code is excluded and must be available separately to rerun a project.
No saved series in this browser.
- Resources
- 2q · 2g
- Compiled depth
- 3
- Noisy success
- 94.9%
- Ideal/noisy TV distance
- 5.1%
Lane comparison
Every basis state shows probabilities from the same preset and seed across the available execution lanes.
|00>|01>|10>|11>Versioned experiment artifact
- Plan identity
- nmplan-0dc681af88ba01d4
- Requested backend
- Auto-select local simulator
- Effective backend
- Mixed local engines
- Ideal lane backend
- Local stabilizer
- Noisy lane backend
- Local statevector
- Compiled gates
- 2
- Measurements
- 2
- Success support
- |00⟩, |11⟩
Evidence boundary
- Histograms are finite-shot samples, not exhaustive probability proofs.
- Every effective backend is a local simulator; no QPU executed the circuit.
- Noise probability is a model input, not a device calibration snapshot.
- Contract 0.1 accepts the published preset circuits only; arbitrary source execution stays in Playground.
Inspect generated noisy N/M source
module experiment_bell_depolarizing;
@target("browser-stabilizer");
@seed(42);
@noise("depolarizing", 0.05);
@mitigation("zne");
fn main() {
let q: QReg<2> = qreg[2];
sample 512 {
H(q[0]);
CNOT(q[0], q[1]);
let a: Bit = measure(q[0]);
let b: Bit = measure(q[1]);
return (a, b);
}
}
Saved Experiment Studio runs
Keep privacy-minimized evidence locally, explicitly sync it to your account, restore its plan, or compare two saved versions.
Select 0/2 saved runs to compare their exact inputs and evidence.
Up to 50 integrity-checked records are retained. Source text is never stored in this history or uploaded to Supabase; only plan inputs, evidence, and source hashes are kept.
RFC roadmap for this surface
These cards expose planned programs. A proposed RFC is not an executable capability or implementation approval.
NM-RFC-0018ProposedImplementation lockedDensity-matrix execution and Kraus channels
- Proposed activation
- experimental.densityMatrixNoise=true
NM-RFC-0019ProposedImplementation lockedCalibration-aware timing and idle noise
- Proposed activation
- experimental.calibrationAwareTiming=true