RESEARCH BLOG • CAMPAIGN POST-MORTEMS

Scientific Deep Dives & Campaign Post-Mortems

Open science requires publishing positive demonstrations, negative findings, and mathematical audits with complete transparency.

LATEST RESEARCH REPORTAUGUST 17, 2026HARDWARE: IBM QUANTUM (ibm_marrakesh — 156Q)

Q-Psi Compiler Reaches Its First Quantum-Advantage Milestone

Subtitle: Nine compiler-derived search cases, three problem sizes, one frozen physical-QPU experiment, and an independent audit.

When Q-Psi initially evaluated the State-Space Compiler on IBM Quantum hardware in our Stage-6F campaign, we confirmed physical-QPU interoperability: our QUBO and Ising formulations successfully mapped onto 156-qubit Heron architecture and recovered exact classical global optima on 100% of small problem instances (N ≤ 10). However, we explicitly did not claim quantum advantage at that time. NISQ gate depth and device noise degraded larger instances, and compiler-specific speedup remained unproven.

To rigorously test whether compiler-generated state spaces can achieve a bounded quantum advantage, we designed a candidate-state search experiment combining the Q-Psi compiler with Grover's amplitude amplification.

THE SCIENTIFIC PROCESS: FROM FLAW IDENTIFICATION TO PHYSICAL VERIFICATION
  1. Pre-QPU Fairness Review: An adversarial audit of protocol v1.0 revealed that candidate ordering correlated with compiler internal rankings. If left uncorrected, classical algorithms could exploit this side channel.
  2. Protocol v1.1 Correction: We redesigned the protocol before submitting QPU jobs, applying answer-independent candidate permutations and an opaque black-box verifier contract with 100.0% mechanical equivalence.
  3. Single Physical Execution: The frozen protocol was executed on ibm_marrakesh (Job ID: da1c7rkdedkc73eqs5mg, 18,432 physical shots).
  4. Unanimous Advantage: 9 out of 9 individual cases and 3 out of 3 problem sizes (N=4, 8, 16) beat the classical expected query cost.
  5. Statistical Rigor at N=16: The 95% Wilson confidence interval upper bound on quantum effective queries (\(7.360\)) remained strictly below the classical expectation (\(8.5\)).
  6. Independent Audit: An independent adversarial audit verified the raw evidence seal (SHA256: 8f65edbe...ff2) and delivered a final verdict of SUPPORTED.

We maintain strict scientific boundaries: this experiment demonstrates quantum query advantage under a black-box verifier model. It does not claim wall-clock runtime speedup or universal software-repair supremacy over all structure-aware classical heuristics.

HISTORICAL RESEARCH REPORTAUGUST 17, 2026HARDWARE: IBM QUANTUM (ibm_marrakesh)

From Software-Repair Compilation to a Physical-QPU Quantum Advantage Experiment: What Q-Psi Tested, What Worked, and What Failed

By the Q-Psi Research Team

A comprehensive overview of the four initial physical QPU campaigns on IBM Quantum's 156-qubit Heron processor, ibm_marrakesh: state-space software repair compilation (Stage-6F), single-shot dynamic Bernstein-Vazirani query advantage, constant-depth restricted Simon circuits, and exploratory mantra quantum state encoding.