Quantum evidence

Claims should be harder than demos.

We preserve the source version, circuit hashes, seeds, telemetry and statistical boundary behind each quantum decoding result. A working interface is not presented as proof of superiority.

Inspect the live system
Recorded evidence

600 million shots. 24 frozen reproductions.

The complete reproduction campaign covers every Beam8, Beam32 and Beam64 configuration across balanced X and Z workloads, with 25 million shots in each cell.

600M+

recorded simulation shots

24 balanced cells across Beam8, Beam32 and Beam64, four seeds, X and Z. The campaign includes exploratory and controlled subsets.

10K

same-syndrome compatibility cases

5,000 X and 5,000 Z diagnostic inputs matched the released reference logical outputs.

80K

controller confirmation cases

A fixed continue-BP policy was checked separately from the beam-width routing claim.

Research programme

Every study has a place in the record.

Published research and reproducible simulation.

Relay BP

Memory strength selection for bivariate bicycle qLDPC codes.

Large scale benchmark evidence with telemetry.

Beam Search

Bounded Beam8, Beam32 and Beam64 decoding campaigns.

Research stage controller and unseen input evaluation.

Adaptive Controller

Predict when extra decoding compute is worth its cost.

Evidence infrastructure used across every campaign.

Validation

Source locks, circuit hashes, seeds, confidence intervals and compatibility checks.

Campaign visualisation

From syndrome to evidence.

Animated view of the measured workflow: decode, allocate compute, record telemetry, then apply the claim gate.

Claim gate

What still separates evidence from a market claim.

Freeze one immutable implementation, run balanced unseen X/Z inputs, compare named baselines on identical workloads, and report reliability plus compute with confidence intervals.

01Immutable source
02Unseen seeds
03Balanced X + Z
04Named baselines
05Confidence intervals
06Compute telemetry
What the evidence establishes
Hyper Orbits has built a reproducible quantum decoding research stack and measured it at scale. The stack combines belief propagation, bounded Beam Search, adaptive compute control, telemetry, provenance and validation. More than 600 million simulation shots, balanced X and Z workloads, compatibility checks and three beam configurations form the evidence base for the next held out evaluation.