What we tested, and what we learned. Misses included.
For each pre-registered experiment in this index, the pass rule was written down and sealed by hash before the run. Each ran once and is reported as-is. The first public cohort (2026-10-13) listed nine pre-registered experiments plus one pipeline test; two additional pre-registered experiments were added on 2026-10-20 (see results[] and preregistered fields). A pre-registered public-mindweb link check dated 2026-10-31 is also in results[]. It reports two labels at the same rank: GNN_BETTER against neighbour counts, and TYPE_PRIOR_BETTER for a type-only prior. It is not an Aurion product claim. A pre-registered public-mindweb link check dated 2026-11-02 is also in results[]. The headline is the pre-registered experiment verdict NO_DIFFERENCE. The diagnostic verdict is NO_DIFFERENCE at the same rank. A gate+gate pair label does not replace that headline. It is not an Aurion product claim. A recorded pair of small quantum tests dated 2026-11-02 is also in results[]. Both recorded verdicts are PASS. The learned-rule pass was expected, so that test is not a stress test. The image result does not certify entanglement. There is no advantage claim. A recorded pair of small quantum checks dated 2026-11-02 is also in results[]. The fidelity verdict is PASS and the readout verdict is IMBALANCE. The fidelity score is raw. It is not certified entanglement, and it is not error-mitigated. The readout matrices were not applied to the fidelity counts. A pre-registered comparison dated 2026-11-02 is also in results[]. The verdict is FAIL. It covers this method on these qubits. It is not a device certificate. A pre-registered comparison dated 2026-11-03 is also in results[]. The verdict is FAIL. It covers this method on these qubits. It is not a device certificate. A pre-registered comparison dated 2026-11-03 is also in results[]. The verdicts are P4-A PASS and P4-C FAIL. It covers this method on these seven qubits. It is a measured result, published pass or fail.
A diffusion generator trained on 300 checkpoints of a small MLP's last layer was asked to produce new weights at equal accuracy. It copied: for every one of 600 generated samples, some training checkpoint predicts the same class on all 450 test inputs.
Read the result and request the full write-upWith symmetry augmentation the generator stopped copying but degraded: mean accuracy 0.86-0.88 versus 0.96 for the checkpoints, and its samples sat about as far away as plain noise.
Read the result and request the full write-upA template search over optimizer update rules won on its training tasks but beat tuned Adam on only 0, 1 and 0 of 4 held-out task families across three seeds (3 of 4 needed). The best rules were bounded Adam variants that overfit two training instances.
Read the result and request the full write-upWith 8 training instances per family the search again failed on held-out families (0, 0 and 1 of 4). The hypothesis that more training instances shrink the train/held-out gap was not supported: it shrank on 2 seeds and grew on 1.
Read the result and request the full write-upTurning a server's prompt cache on changed 39 of 720 replies (5.4%). Strict verdicts changed in 5 of 720 cases, which is not significant (the confidence interval starts at 0).
Read the result and request the full write-upOn a randomly initialised tiny model both methods sat near chance; the score ratios were 0.842, 1.059 and 0.957 (0.9 needed on every seed). Inconclusive.
Read the result and request the full write-upFrom a pretrained tiny base, AdamW-LoRA reached 0.987-0.995 while MeZO reached 0.55-0.60 of that score (0.9 needed). This pilot separates the methods; MeZO scored almost nothing on the generation task.
Read the result and request the full write-upA pre-registered per-task viability gate was applied to a small adapter-training pilot on six tasks. Only 3 of 6 tasks were viable (the rule needed 4), so the gate said stop.
Read the result and request the full write-upOn 64 held-out graphs (CPU simulation), a learned model wrote QAOA angles with a mean approximation ratio of 0.766 versus 0.812 for one fixed angle set. The pilot is valid; the learned generator missed.
Read the result and request the full write-upA 2-qubit entangled circuit on Rigetti's 108-qubit Cepheus chip (via Amazon Braket) matched on 98 of 100 shots; the simulator matched 100 of 100. A pipeline test, not a science result; 100 shots can estimate this run's observed mismatch proportion only with wide uncertainty, not a general device or circuit error rate.
Read the result and request the full write-upQMAI evaluated a pre-registered readout statistic in two valid run blocks on the Cepheus-1-108Q device. The A-arm primary test rejected in B1 and B2 at the fixed per-block threshold of 1/300, with exact p-values of 8.68e-12 and 7.49e-14 and positive excess statistics of +0.00213 and +0.00257. A third required block was not run because of the queue gate (the device's job queue held 2 other jobs at the block's start time, above the pre-set limit of 0, so no task was created), so the campaign disposition is INCOMPLETE. No replication or non-replication conclusion is available. This note reports the two per-block results, the missing block, the recorded recomputation, and the limits that remain.
Read the result and request the full write-upQMAI ran a pre-registered, three-resolution check of a three-body gravity scene driven by eight seeded player commands. The checker first passed three controls, then tested four seeds chosen before the episodes. All four seed runs emitted zero flags; their eight energy-error ratios ranged from 3.9999905410799816 to 3.999997730513573, inside the frozen inclusive band from 3.96 to 4.04. The largest momentum or angular-momentum residual in the seed runs was 1.1102230246251565e-15, and every such residual stayed below its pre-registered roundoff bound. Under the frozen rule, the report's verdict is G1_PASS. This runs on a test branch of our engine and is not in a released game yet; there are no collisions in this world yet.
Read the result and request the full write-upExperiment GNN-0 is reported as recorded. Two labels were frozen before the run and are given the same rank. The primary verdict is GNN_BETTER: graph-model test AUC 0.922737 against the best neighbour heuristic 0.540323, difference 0.382414, 95 percent interval 0.312695 to 0.44719. The diagnostic verdict is TYPE_PRIOR_BETTER: a type-only prior scored 0.993757, difference -0.07102, interval -0.136446 to -0.015479. Only 7 of 171 node-type pairs share an edge, and gate+gate plus doc+memory are 531 of 629 edges. Not an Aurion product claim.
Read the resultExperiment GNN-1 is reported as recorded. The headline is the pre-registered experiment verdict, NO_DIFFERENCE. Text-GCN AUC minus untrained text-cosine AUC has mean 0.072784, 95 percent interval -0.028017 to 0.172305, across 2 pairs. The diagnostic verdict is also NO_DIFFERENCE. A GNN_BETTER label on gate+gate does not set the experiment verdict. Not an Aurion product claim.
Read the resultTwo small tests ran on four qubits of IBM's 156-qubit ibm_fez processor on 2026-11-02. The learned-rule test passed: 99.5% of shots landed on the 4 predicted patterns (95% CI 99.2-99.8%), and the control put 26.5% of shots on those patterns. The image test passed: total variation 0.178 for the entangling circuit and 0.554 for the product circuit. The learned-rule pass was expected, so that test is not a stress test. The image result does not certify entanglement. There is no advantage claim.
Read the resultTwo small checks ran on three qubits of IBM's 156-qubit ibm_fez processor on 2026-11-02. The fidelity witness passed its raw threshold: F_raw is about 0.91, with a 95% interval of about 0.83 to 0.98. The score is raw. It is not certified entanglement, and it is not error-mitigated. The readout check on the same qubits returned IMBALANCE, from qubit 29 alone. Those assignment matrices were not applied to the fidelity counts.
Read the resultOn IBM's 156-qubit ibm_fez, a measure-and-correct preparation of a 7-qubit GHZ state scored below a CNOT ladder: the pre-registered difference interval is about -0.30 to -0.13, entirely below zero. FAIL for this method on these qubits; it is not a device certificate.
Read the resultpre-registered; this method on these qubits; not a device certificate.
Read the resultpre-registered; this method on these seven qubits; measured result, published pass or fail
Read the resultAn inbound-mail replacement candidate passed isolated tests; production remains unchanged.
Read the updateAn AI agent delivered notebook code that passed all 177 cases of an acceptance suite written before the code. Run as shipped, on fake data, the notebook's code cell stopped with a TypeError at line 211 in each of three runs. The suite's cases had run that cell only as far as line 195.
Read the articleOn one Windows 11 machine, we measured two or more renames of the same file all reporting success. With five processes released together, that happened to every one of 600 files in one run and to 599 of 600 in a second.
Read the articleOn 10 November 2026, the instruction for redeploying a web app that runs on credits and takes no payment online said to copy the two changed files into a new container. The agent that deploys created the container without starting it and hashed all seven app files against the manifest of the reviewed change: five matched and two did not. It paused with the running site untouched, the instruction was corrected to copy all seven files, and the switch from the old container to the new one then took eleven seconds.
Read the articleWhat we tested, and what we learned (90 s)
Video hash: 3de55bc330bba5f8
Full write-ups are not public yet. You can request one; requests are reviewed by hand.
Content inventory: FINDINGS_INVENTORY_20261014_r4.md
Inventory hash: c5b7ebbf9cca5da3