2026-03-18 · build 29 · Bell State |Φ+⟩CLAIM CORRECTED
First Contact: Hello IBM (claim corrected)
A Bell-state circuit submitted to IBM hardware; the published value was computed on a simulator (see the correction above). The prior "fidelity 1.0000 / zero noise" value is retracted. Perfect fidelity and zero noise are physically impossible on NISQ hardware; A later hardware run on 20 March 2026 recorded ≈0.9531. Real execution happened; the perfect-fidelity claim did not.
IBM job ran · published value computed on a simulator (see correction)
- Bell Fidelity (hw)
- ≈0.95
- Claim corrected
- 2026-05-15
- Prior false value
- 1.0000 / Noise 0.0000
2026-03-19 · build 35 · CHSH / SATOR HMAC: Bell-certification claim RETRACTEDCLAIM RETRACTED
SATOR HMAC: Bell-certification claim retracted
A CHSH circuit submitted to IBM hardware; the published value was computed on a simulator (see the correction above). The prior "Quantum-Certified SATOR HMAC, authentication requires a live Bell violation (S>2.0)" claim is fully retracted: our run log records the CHSH Bell test REJECTED (S≈0; a Bell violation requires S>2). There is no hardware Bell certification. Defensible: SATOR HMAC remains a real classical authentication primitive, and the CHSH-gate is a protocol design, but it was not certified on quantum hardware.
IBM job ran · published value computed on a simulator (see correction)
- Status
- CLAIM RETRACTED
- Hardware CHSH
- REJECTED (S≈0; violation needs S>2)
- Prior false claim
- live Bell violation gates HMAC signing
- Corrected
- 2026-05-15
2026-03-19 · build 36 · E8 Quantum Walk: 8-Qubit Trotterized Cartan EvolutionCLAIM RETRACTED
E8 Quantum Walk: α detection claim retracted
An 8-qubit Trotterized E8-Cartan walk ran as an IBM job (published value computed on a simulator, see correction). The prior "fine-structure constant α=1/137 directly detected (Δ=0.021 rad)" claim is fully retracted. The artifact recorded alpha_detected:true while the alpha_hits list was empty ([]). θ_α=π/137 is an input design angle, not a measured result. Defensible: a real-hardware Trotterized E8 walk; the Coxeter sub-claim is unverified pending its own null.
IBM job ran · published value computed on a simulator (see correction)
- Qubits
- 8 (E8 rank)
- Coupling (input)
- θ_α = π/137 per step (design parameter)
- Status
- CLAIM RETRACTED
- Prior false claim
- α=1/137 directly detected (Δ=0.021 rad)
- Artifact reality
- alpha_detected:true while alpha_hits:[] (empty)
2026-03-19 · build 38 · E8 QPE: α=1/137 detection claim RETRACTEDCLAIM RETRACTED
E8 QPE: α=1/137 detection retracted (real-hardware QPE only)
A 15-qubit QPE circuit (128 bins) ran as an IBM job on an older 133-qubit machine; the job ID is genuine, but the published value was computed on a simulator (see correction). The prior "α=1/137 detected as a distinct eigenphase, SNR 25.4×, reduction to practice" claim is fully retracted. Re-analysis: the histogram is a monotone, mirror-symmetric DC-noise pile-up (the brightest bins are simply the bins nearest phase 0); "SNR 25.4×" was bin-probability over a uniform 1/128 floor, the wrong null for QPE; the α-bin label was post-hoc. No α eigenphase was resolved. Defensible: real-hardware QPE execution with a public job ID.
IBM job ran · published value computed on a simulator (see correction)
- Precision
- n=7 (15 qubits, 128 bins); job ran, value from simulator
- Status
- CLAIM RETRACTED
- Prior false claim
- α at bin 1, SNR 25.4×, distinct eigenphase
- Re-analysis
- DC-noise pile-up: top bins = the bins nearest phase 0; 'SNR' = bin-prob ÷ uniform 1/128 (wrong null)
- Mirror 'coherence'
- the mirror symmetry IS the pile-up signature, not coherence
2026-03-19 · build 40 · E8 Gauge Decomposition: Standard Model Subgroup ProjectionsUNRESOLVED: INSUFFICIENT EVIDENCE
E8 Gauge Decomposition: unresolved (α: insufficient evidence)
Three IBM jobs (published values computed on a simulator, see correction), same input angle θ_α=π/137, different graph topology. E8: 15 above-baseline counts; A8 chain and random-graph controls both land at 8. Corrected 2026-07-03 after independent forensic re-derivation: entropy and KL-divergence-from-uniform are statistically indistinguishable across all three circuits (E8 7.392/8 bits, A8 7.388/8, random 7.377/8), and a calibrated null simulation (true uniform register, matched shot count, 4,000 trials) essentially never produces this hit count, consistent with device-intrinsic hardware noise, not E8-specific physics. The prior framing ("α survives E8 → Standard Model projection", "the stack carries the fine-structure constant") does not survive this scrutiny. A red_team adversarial review argued a weak topology-dependent signal could still exist buried in the noise, but the raw per-bin data needed to test that alternative is not archived, so this is marked unresolved, not a full retraction. Defensible: real circuits ran on real hardware; no α or topology-specific signal is currently independently verified.
IBM jobs ran · published values computed on a simulator (see correction) · 3 circuits
- E8 (full)
- 15 above-baseline counts
- A8 chain control
- 8 (baseline)
- Random graph
- 8 (same baseline)
- Entropy/KL
- statistically indistinguishable across all three circuits
- Honest finding
- unresolved, insufficient evidence for a topology-specific signal