Quantum Research — the science of quantum in physics & life

Behaviour is replication-grade; mechanism is not. In pigeons alone 2025-26 produced a vestibular candidate, a hepatic candidate and a retired beak candidate, while cryptochrome remains the songbird favourite on in-vitro and computational grounds (Luo et al., JACS 147:43934, Nov 2025 — verified). Lam & Malkemper (Physiol Rev 2026) call the field 'the unsolved mystery'.

★ For the kids: Birds definitely have a magnetic compass. Which body part does it is still unknown — and the suspect list got longer this year.

Technical: Watch-items: any causal-with-rescue result; any two-lab receptor replication; any Cry4 loss-of-function behaviour in a bird.

Frozen prediction: By 2028-09-05 no Cry4 loss-of-function bird shows a peer-reviewed magnetic-compass deficit (p=0.75).

Held at modeled. The proposers were right that the standing annotation ('knockout evidence, single research lineage', referring to monarch cryptochrome-1) is out of date, and wrong about what replaces it. The termite work is a genuine second lineage for insect magnetic ORIENTATION, but it is not a second lineage for the radical-pair reading: the same laboratory's 2021 Communications Biology paper concluded that magnetic particles with an olfactory co-receptor matter alongside Cry2. So the corrected annotation reads: two lineages, neither rescue-controlled, and the second lineage's own publish…

★ For the kids: We used to have only butterflies as evidence that insects sense magnetism with a light-sensitive protein. Now termites are a second case — but the same termite lab also said tiny magnetic particles are involved, which is the non-quantum explanation.

Technical: Wang W et al., Insect Mol Biol (2026), doi:10.1111/imb.70060 (verified). Prior same-lab result: Gao Y, Wen P, Cardé RT, Xu H, Huang Q, Commun Biol 4:1121 (2021), doi:10.1038/s42003-021-02661-6. Standing monarch lineage unchanged. Honeybee counterweight: bee-magnetite-compass (open) plus the Haase imaging platform (preprint, open). Promotion to viable requires a rescue-controlled manipulation in at least one insect — the evidence class that is empty across this entire lens.

Frozen prediction: Carried by termite-nos-magnetic-orientation.

Held at open; two proposers filed overlapping components (drosophila-magnetosensitivity, drosophila-cry-in-vitro-mfe-paradox) resting on the same primary paper, merged here so one result is not counted twice. The 2025 molecular evidence: HDX-MS on purified DmCRY found a reversible, long-lived blue-light conformational switch in the C-terminal tail with NO detectable magnetic-field effect on that switch at 18-25 mT — a field roughly 400x geomagnetic — and the same work reports that a tryptophan-chain-disrupted mutant shows a far larger radical-pair magnetic-field effect than wild type. Read ca…

★ For the kids: Scientists shone light on a fly protein and watched it change shape, then added a strong magnet — the shape change ignored the magnet. Oddly, a broken version of the protein noticed magnets more than the working one. Nobody tested actual flies in that study.

Technical: Chandler SA, Gehrckens AS, Shah LMN, Buckton KE, Cao G, Sen N, Zollitsch T, Solov'yov IA, Schleicher E, Weber S, Hore PJ, Timmel CR, Mackenzie SR, Benesch JLP, 'Light-induced conformational switching and magnetic sensitivity of Drosophila cryptochrome', Structure 33(11):1930-1943.e4 (2025), doi:10.1016/j.str.2025.08.003 — verified (PubMed 40882633). The authors frame the null as 'too small to detect under the conditions of the HDX experiment', so it bounds rather than excludes an MFE, and it does NOT test the Bradlaugh FAD-less C-terminal-52-residue construct. Behavioural null: Bassetto et al…

Frozen prediction: By 2027-08-25, P=0.85 that no independent laboratory (no Manchester/Leicester/Oxford co-authors) publishes a peer-reviewed behavioural magnetic-field effect in Drosophila with n≥1,000 and pre-registered analysis. If one appears, promote to modeled; if none, retire per the frozen clock.

Pantry lead claim:kraken:ed52289d28fd chased to its real source and VERIFIED: Zhou, Chen, Yuan, He, Tiwari, Gelin, Remacle, Miller, Zheng, Jha, Duan, 'Environmental Control Extends Beyond Quantum Dephasing in Exciton Energy Transfer', arXiv:2608.23423 (24 Aug 2026); the abstract carries the lead sentence verbatim. Temperature-dependent 2DES on allophycocyanin 10-296 K: beta->alpha transfer time 400 fs (10 K) -> 200 fs (30-40 K) -> 400 fs (296 K) while dephasing falls monotonically; HEOM fits need an anharmonically temperature-evolving low-frequency spectral density. Consistent with the corpus…

★ For the kids: Sunlight energy hopping inside a light-harvesting protein is real. A new (unchecked) measurement says the protein's own wiggle pattern sets the pace, fastest at one cold temperature.

Technical: Falsifiers: independent 2DES reproduction of the 30-40 K minimum; a rival-model fit (vibronic-exciton vs Forster/modified-Redfield with T-dependent lineshapes) — if both fit, 'functional coherence' stays an interpretation.

Frozen prediction: By 2027-09-05 arXiv:2608.23423 is published in a peer-reviewed journal (p=0.7); by 2028-09-05 an independent laboratory's temperature-dependent 2DES on allophycocyanin reproduces a beta->alpha transfer-time minimum between 20 K and 60 K (p=0.3).

The theory that smell reads molecular vibrations via inelastic electron tunnelling. The keystone evidence - human H/D isotope discrimination - did not survive scrutiny: Block et al. (PNAS 2015) showed the human musk receptor OR5AN1 (and others) fail to distinguish isotopomers in vitro and argued the tunnelling step is biophysically suppressed. Kept OPEN (lowest tier), flagged NEAR-RETIREMENT; the specific 'isotopes prove it' keystone is retired (see retired[]).

★ For the kids: Maybe the nose senses a molecule's tiny 'wiggles' with a quantum trick. The main proof was people smelling heavy vs normal versions of a molecule - but a careful test found the smell receptor can't tell them apart. So this idea is mostly out.

Technical: Block et al. (PNAS 2015) receptor-level + theoretical rebuttal of Turin's vibrational/swipe-card model; inelastic-electron-tunnelling readout easily quenched and inconsistent with combinatorial shape-based OR coding. Shape theory remains mainstream.

Frozen prediction: Within the horizon no published human-psychophysics + receptor study reverses Block et al. (2015) to re-establish vibrational olfaction.

The 2026 tubulin isotope paper is spin chemistry on a protein assay: no coherence in a neuron, no cognitive readout, no test of orchestrated reduction. Firewall maintained; barred from rising regardless of publication volume.

★ For the kids: A chemistry result about cell scaffolding is not a result about minds.

Technical: Awareness claims require a cognitive or neural-dynamics readout co-varying with a manipulated quantum variable, in an animal, replicated. None exists.

Frozen prediction: By 2028-09-05 no peer-reviewed study links a manipulated nuclear-spin variable in tubulin to a cognitive or anaesthetic endpoint in an animal with independent replication (p=0.9).

VERIFIED and CORRECTED: Bluvstein et al., 'A fault-tolerant neutral-atom architecture for universal quantum computation', Nature 649:39-46 (Jan 2026, online Nov 2025; doi:10.1038/s41586-025-09848-5) — up to 448 atoms, Lambda = 2.14(13) in a FOUR-ROUND characterization circuit with loss detection + ML decoding (proposal 2 cited it as 'Nature 2025'). Google Willow, Nature 638:920 (2025) — Lambda = 2.14 with real-time decoding (the identical number is a coincidence; both verified). VERIFIED: Paetznick et al. (Microsoft Quantum + Quantinuum), Nature 654:349 (Jun 2026; doi:10.1038/s41586-026-10628…

★ For the kids: Three kinds of quantum computer now correct errors well enough to beat their raw parts in tests — but one of them throws away runs where it caught a mistake, and none of them counts the moments when a space particle hits the whole chip.

Technical: Post-selection acceptance rates are the number to read before comparing Paetznick to memory experiments. Promotion needs a sustained (>= 1 h, burst-inclusive) Lambda >= 2 across three distances on two platforms.

Frozen prediction: By 2027-09-05 no trapped-ion paper reports a distance-scaling factor Lambda >= 2 across three code distances without post-selection (p=0.7).

NIST finalized its first three post-quantum cryptography standards in August 2024 (FIPS 203 ML-KEM/Kyber, FIPS 204 ML-DSA/Dilithium, FIPS 205 SLH-DSA/SPHINCS+), converting a decade of theory into a mandated migration target.

★ For the kids: The people in charge of US computer-security rules just finalized the actual new 'quantum-proof' secret-code recipes — so companies can start really switching over, not just planning to.

Technical: NIST FIPS 203/204/205 (August 2024) standardize lattice-based KEM/signature schemes and a hash-based signature as approved algorithms; paired migration guidance (CNSA 2.0) sets phased deadlines through 2033.

Frozen prediction: At least one G20 government publishes a binding (not merely advisory) PQC migration deadline earlier than 2030 within the horizon.

VERIFIED: Kavokin, Bojarinova, ... Chernetsov, J R Soc Interface 23:20260129 (Jul 2026); Rybachy, 2023-2025 seasons; 1.41/1.5 MHz carriers, 500 Hz square-wave AM. Modulated field at half mean power disoriented pied flycatchers at lower carrier amplitude than CW — opposite to the mean-power scaling of cryptochrome decoherence. One group with a long prior position. RF attribution to the RPM stays contested; the phenomenon of RF disruption is not.

★ For the kids: Chopped-up radio waves confused the birds more than steady ones with more power. If the compass were a tiny spin sensor it should be the other way round.

Technical: Confounds: the 500 Hz envelope adds low-frequency components an induction sensor could read directly; nonlinear RPM responses (saturation, Zeno-type) not excluded. Leberecht et al., PNAS 2023 remains the broadband upper bound.

Frozen prediction: By 2027-09-05 a group with no Kavokin co-authors publishes a matched-mean-power modulated-vs-CW RF test in a night-migratory bird (p=0.35).

VERIFIED: Luo, Hungerland, Solov'yov, Subotnik, Hammes-Schiffer, JACS 147:43934 (Nov 2025); follow-up on the protein–solvent interface, JACS 148 (2026). These refine parameters; they do not test the model.

★ For the kids: Computer models say the compass protein keeps its separated electron pair stable thanks to the water and protein around it — a better model, not a new measurement.

Technical: Falsifier: time-resolved EPR/MARY on Trp mutants discriminating the composite Trp3/Trp4 pair from canonical FAD–Trp4.

Frozen prediction: By 2027-09-05 no experimental (EPR/MARY) test discriminating the composite Trp3/Trp4 radical pair from the canonical FAD–Trp4 pair in Cry4a is published (p=0.6).

Held at viable, with the scope narrowed exactly as the corpus already narrowed photosynthetic coherence — but the split is executed differently from what the proposers filed. What is replication-grade: hydrogen transfers in enzymes proceed by tunnelling, evidenced by anomalously large kinetic isotope effects with temperature dependences classical over-the-barrier transfer cannot produce, reproduced across many enzymes and many laboratories over three decades and resolved mechanistically this year in plant lipoxygenases. That is the strongest quantum-effect-in-biology claim the corpus carries …

★ For the kids: Enzymes really do use a quantum shortcut where hydrogen slips through a barrier instead of climbing over it. That part is solid and lots of labs have shown it. Whether the enzyme was built to USE that shortcut is a separate question we moved to 'unknown'.

Technical: Occurrence leg: Protein Sci 35:e70647 (2026), doi:10.1002/pro.70647, rate-limiting hydrogen tunnelling in plant 9- and 13-lipoxygenases; Biochemistry (2026), doi:10.1021/acs.biochem.5c00769; J Phys Chem B (2026), doi:10.1021/acs.jpcb.6c02719, donor-acceptor separation augmenting KIE temperature dependence in NADH-model hydride transfer. The distinction the corpus enforces everywhere: a quantum effect being PRESENT is a different claim, at a different tier, from that effect doing biological WORK.

Frozen prediction: P<0.05 by 2030 that the anomalous KIE temperature dependences themselves fail to reproduce. That, and only that, would falsify this component's viable tier.

Lisowski et al. (Science 2026) keep the iron hypothesis alive in a new form. The map-sense component stays at modeled on behavioural displacement evidence in multiple species; the receptor side inherits the openness of pigeon-liver-macrophage-magnetosense.

★ For the kids: The 'iron in the body' compass idea is not dead — it moved from the beak to the liver, and now needs checking.

Technical: Superparamagnetic particles hold no remanence; torque-based transduction requires converting field-dependent susceptibility into a cellular signal — unspecified.

Frozen prediction: By 2028-09-05 no independent lab reproduces a magnetometry signal in pigeon liver macrophages exceeding that of any other tissue (p=0.55).

A 2016 claim that an iron-sulfur protein complex (MagR/IscA1, with cryptochrome) is a rod-like magnetoreceptor. Physically challenged as ~5 orders of magnitude too weak to beat thermal noise at body temperature, and lacking independent replication. Correctly OPEN/contested - a cautionary over-claim, pinned here and guarded against any rise (see retired[]).

★ For the kids: One team said they found the actual 'compass protein.' Others did the physics and said it's far too weak to feel Earth's magnetism through all the warm jiggling - so this stays a big maybe.

Technical: Qin et al. (Nat Mater 2016) proposed a conserved MagR/Cry rod complex; Meister (eLife 2016) showed the Zeeman coupling sits ~10^5x below thermal fluctuations at 300 K. No unaffiliated replication; treated as contested.

Frozen prediction: Within the horizon MagR/IscA1 is NOT established as a physiological magnetoreceptor; Meister's thermal-noise objection stands.

DEMOTION AFFIRMED at open. This component and 'cellular-autofluorescence-mfe' rest on the SAME single paper, yet the corpus carried the specific result at open and the general claim one tier higher. A generalisation cannot outrank the only observation supporting it. No independent replication, no second cell line, no Zeeman-matched RF suppression control, and the in-cell partner radical (superoxide vs Trp) is itself the physically marginal branch.

★ For the kids: One lab saw a magnet slightly change the faint glow of ordinary human cells. We'd written that up twice — once carefully and once too confidently. Now both say the same thing: interesting, unrepeated.

Technical: Ikeya & Woodward, PNAS 118, e2018043118 (2021): ~3–4% modulation of HeLa flavin autofluorescence under weak static fields. Framework review: Zadeh-Haghighi & Simon, J. R. Soc. Interface 19, 20220325 (2022) — a framework paper shows RPM CAN accommodate reported effects, not that any effect REQUIRES a spin-correlated pair. Superoxide-pair viability: Player & Hore, J. Chem. Phys. 151, 225101 (2019).

Frozen prediction: Promotes back to modeled only on independent-lab replication in a second cell line PLUS a Zeeman-matched RF suppression control, by 2028-12-31.

SCOPE NARROWED, tier held. This component now covers ONLY (a) the single-photon electrical response of rod photoreceptors and (b) ultrafast cis–trans photoisomerisation of retinal through a conical intersection in ~50–200 fs at near-unity quantum yield. Both are measured, reproduced across four decades and mechanistically specific — one of the strongest quantum-biology cases that exists. The human PERCEPTUAL claim has been removed and demoted separately; a component mixing a forty-year-replicated biophysics result with a marginal unreplicated psychophysics result launders the tier of the weak…

★ For the kids: One light particle really can flip a molecule in your eye and make a cell fire — measured, re-measured, true. We just moved the harder question ('can a PERSON notice one photon?') into its own box, because it's much shakier and was borrowing the strong box's badge.

Technical: (a) Baylor, Lamb & Yau, J Physiol 288:613-634 (1979) — suction-electrode recordings, quantised photocurrent. (b) Polli et al., Nature 467:440-443 (2010) — femtosecond pump-probe resolving barrierless conical-intersection passage; the non-adiabatic surface crossing is directly observed, not inferred. Neither layer requires or implies long-lived coherence, entanglement, spin, or any awareness claim.

Frozen prediction: No demotion trigger outstanding for layers (a) and (b); re-audited only if a conical-intersection reassignment analogous to the vibronic reassignment of photosynthetic coherence is published.

A concrete, intriguing result: xenon isotopes carrying nuclear spin are less potent anesthetics than spin-zero isotopes in mice - pointing to spin-selective (radical-pair-like) chemistry in anesthetic action. FLAGGED and firewalled: this is spin-CHEMISTRY/pharmacology, explicitly NOT evidence of quantum 'awareness'/Orch OR. Open pending replication.

★ For the kids: Different 'flavors' of the gas xenon put mice to sleep a little differently depending on a quantum property (nuclear spin). A real quantum-chemistry clue - but NOT proof that thoughts are quantum. We keep those ideas apart.

Technical: Spin-carrying 129Xe/131Xe less potent than spin-zero 132Xe/134Xe (Li et al., Anesthesiology 2018), consistent with hyperfine modulation of a radical-pair step. Adjacent to - but NOT - the Orch OR awareness claim, which stays open and unasserted.

Frozen prediction: Within the horizon the xenon nuclear-spin anesthetic effect, if replicated, is interpreted as spin-selective radical-pair pharmacology, NOT as evidence of quantum awareness; the Orch OR line remains open and unasserted.

Loewdin's old idea - a proton tunnels across a DNA hydrogen bond to create a rare tautomer that mispairs and seeds a point mutation. Recent open-quantum-systems modelling finds non-negligible tautomer populations, but a MEASURED causal contribution to real spontaneous-mutation rates is unestablished. Genuinely open/speculative.

★ For the kids: A wild idea: sometimes a tiny particle inside DNA might 'tunnel' to the wrong spot and cause a typo in the genetic code. It's modelled on a computer, not measured in cells.

Technical: Slocombe, Sacchi & Al-Khalili, Commun Phys 5, 109 (2022): open-quantum-systems model of G-C bonds finds tunnelling-driven tautomer inter-conversion; downstream mutational impact not demonstrated. Open: do tautomers survive to replication, and can polymerase-fidelity assays detect the contribution?

Frozen prediction: Within the horizon no replicated study establishes DNA proton-tunnelling tautomers as a measurable contributor to spontaneous point-mutation rates.

HELD AT OPEN, FLAGGED. AWARENESS UNASSERTED AND UNASSERTABLE. Tryptophan mega-network exciton diffusion and UV superradiance are measured, and are now shown to be generic to large tryptophan networks rather than microtubule-special. The firewall is tightened: no ultraweak-photon-emission or biophoton result may be used to support microtubule optical signalling or any Orch-OR-family claim, because UPE is ROS chemiluminescence with no demonstrated coherence or information channel, and the strongest in-vivo neural-biophoton claims were shown by the imaging group itself to be background contamina…

★ For the kids: Tiny scaffolding tubes inside cells really do move light energy around. That's measured and real. It does NOT mean the tubes are thinking, and the faint glow living things give off isn't the brain sending light messages — that turned out to be stray light in the room.

Technical: ADMISSIBILITY RULE effective 2026-08-01: results from ultraweak-photon-emission-biology are inadmissible as support for this domain or any Orch-OR-family domain. The maturity ceiling on awareness claims is 'open' by protocol regardless of what optical measurements show. Generality result: tryptophan-network optics are not microtubule-specific.

Frozen prediction: By 2030-12-31 no publication reports a microtubule optical or excitonic observable whose measured value tracks a behavioural or perceptual state variable in a living organism.

Kerskens & Lopez Perez, J Phys Commun 6:105001 (2022) → Warren Comment (2023; classical iZQC can yield the signal) → Reply (2023). No independent replication as of 2026. VERIFIED: Wiest, Neurosci Conscious 2025 niaf011 abstract describes this as 'direct physical evidence of a macroscopic quantum entangled state in the living human brain' — a counted case for quantum-hype-in-peer-review.

★ For the kids: One team saw a strange heartbeat-timed signal in a brain scanner and called it entanglement. Another expert said ordinary physics could make that signal. Nobody else has seen it.

Technical: iMQC-as-witness assumes any correlation not explainable classically implies a non-classical mediator (LOCC/BMV logic); classical intermolecular dipolar / distant-dipolar-field explanations not excluded; no unaffiliated replication (J Phys Commun Comment + Reply, 2022-2023).

Frozen prediction: By 2027-12-31 no independent group (no Kerskens/Lopez Perez authorship) publishes a peer-reviewed replication of the heartbeat-locked ZQC signal with the Warren long-range-dipolar controls (p=0.80).

Held MODELED. Reconfigurable atom arrays ran algorithms on dozens of error-correcting logical qubits with mid-circuit reconfiguration (Bluvstein et al., Nature 2024) and later fault-tolerant logical operations. Strong milestones, but sustained many-round below-threshold logical memory (error DECREASING under scaling) is NOT yet shown — hence modeled, not viable.

★ For the kids: Lasers hold single atoms like beads on an invisible grid and shuffle them to run error-protected quantum programs.

Technical: 48 logical qubits, transversal gates, tweezer reconfiguration (Bluvstein et al., Nature 2024); logical demos 2024-2025. Open gap: many-round below-threshold logical memory on atoms.

Frozen prediction: A neutral-atom processor demonstrates a logical memory whose error decreases over >=3 rounds under code-distance scaling within the horizon.

Distinct from error-suppression SCALING: running ALGORITHMS on encoded logical qubits. Neutral-atom arrays executed circuits on up to 48 logical qubits (Bluvstein 2024); superconducting bosonic 'cat'/GKP codes passed break-even (Sivak 2023); ML decoders (AlphaQubit, Bausch 2024) improve logical accuracy. Real, early, modality-plural - but far from fault-tolerant scale, so modeled.

★ For the kids: Instead of one fragile qubit, you bundle many into one sturdy 'logical' qubit - and scientists are starting to run small real programs on those.

Technical: Reconfigurable atom arrays with transversal gates (Bluvstein, Nature 2024); GKP bosonic real-time QEC beyond break-even (Sivak, Nature 2023); transformer decoder on Sycamore data (Bausch, Nature 2024); trapped-ion logical qubits (Quantinuum). Overhead and real-time decoding remain the barriers.

Frozen prediction: A logical-qubit processor executes a small algorithm with logical error below that of its physical constituents on a non-superconducting modality within the horizon.

SCOPE DEMOTION, not a tier change. This viable domain now covers UNENTANGLED NV and atomic magnetometry only — the deployed, commercial, replication-grade case at fT/√Hz in fielded instruments with independent replication across many groups. Entanglement-enhanced variants and quantum illumination are carved out and graded separately. A single 'quantum sensing' bucket let first-demonstration and refuted results inherit credibility earned by commercial magnetometers; that is exactly the drift this ledger exists to catch.

★ For the kids: Some quantum sensors really do work and you can buy them — they measure magnetic fields far too small for a normal detector. But not everything with 'quantum' in the name is at that level. We split the shelf so the proven ones don't lend their good name to the unproven ones.

Technical: Barry, J.F. et al. (2016), PNAS 113(49), 14133-14138 — wide-field NV-diamond magnetometry resolving picotesla-scale extracellular fields. NV-diamond magnetometry as a technique class is broadly replicated across many independent groups/applications, supporting viable tier even though this specific single-neuron demo is one paper.

Frozen prediction: NV-diamond (or equivalent quantum) magnetometry directly measures an in-vivo magnetic-field-correlated neural signal in a migratory animal's magnetoreceptive tissue within the horizon.

Strontium/ytterbium optical-lattice clocks reach 10^-18 to 10^-19 fractional frequency - precise enough to resolve gravitational redshift across a millimetre of height (JILA 2022). Replicated across multiple national labs; the clearest case where engineered quantum coherence delivers world-leading MEASUREMENT (not computation). Viable.

★ For the kids: Clocks so exact they can tell that time runs very slightly faster for your head than for your feet.

Technical: Bloom et al. (Nature 2014) 10^-18 clock; Bothwell et al. (Nature 2022) resolved redshift across a mm-scale sample; Ludlow et al. (RMP 2015). Enables relativistic geodesy, a redefinition of the second, and fundamental-constant-drift tests.

Frozen prediction: An optical-clock comparison reaches ~10^-19 fractional frequency (enabling cm-scale relativistic geodesy) within the horizon.

Twin-field QKD has pushed fiber-based quantum key distribution past the 1,000 km mark without trusted relay nodes (Liu et al., PRL 2023) — the practical roadmap toward metro/regional QKD without satellite relays, though cost and integration into existing network infrastructure remain the deployment bottleneck.

★ For the kids: Scientists sent secret quantum keys through a normal fiber-optic cable over 1,000 kilometers without needing a 'trusted middleman' station along the way — a big step toward everyday quantum-secure internet links.

Technical: Liu, Y. et al. (2023), PRL 130, 210801 — a sending-or-not-sending twin-field protocol variant beating the fiber repeaterless (PLOB) bound at long distance.

Frozen prediction: A twin-field QKD link operates as continuous commercial (non-demonstration) infrastructure between two metro hubs within the horizon.

Headline 'quantum supremacy' claims - random-circuit sampling (Google 2019), boson sampling (USTC Jiuzhang 2020) - demonstrated tasks hard for THEN-current classical methods, but several were later matched or spoofed by improved classical algorithms. Correctly OPEN: real physics milestones, contested as durable, practically-useful advantage. Faces the same de-hyping axis as quantum-advantage-claims (NISQ utility).

★ For the kids: Some quantum computers did a weird math task faster than any normal computer - then clever programmers caught the normal computers back up. So 'quantum wins' is still an argument, not a settled fact.

Technical: Arute et al. (Nature 2019) RCS; Zhong et al. (Science 2020) boson sampling; classical narrowing/spoofing e.g. Pan, Chen, Zhang (PRL 2022). No advantage claim yet targets a practically useful problem.

Frozen prediction: At least one pre-2025 random-circuit-sampling 'advantage' headline is matched or spoofed by an improved classical algorithm within the horizon.

The NISQ-era counterpart to below-threshold QEC on the HYPE axis: claims a noisy device did something classically intractable. IBM's 2023 'quantum utility' Nature result (127-qubit kicked Ising) was matched or beaten within months by classical tensor-network and Pauli-propagation methods. OPEN: the hardware is real; the advantage claim is not durable. A facet of the same pattern as quantum-computational-advantage - both kept as the froth control, distinct from genuine capability (below-threshold-qec).

★ For the kids: Companies keep announcing their quantum computer did something no normal computer could. Then, pretty fast, someone does it on a normal computer. The machines are real; the bragging usually isn't.

Technical: Kim et al. (Nature 2023) 'utility before fault tolerance' vs Tindall, Fishman, Stoudenmire & Sels (PRX Quantum 2024, belief-propagation tensor networks) and Pauli-propagation methods that reproduced the observables more accurately. Generalizes prior 'supremacy' walk-backs.

Frozen prediction: Each headline NISQ 'advantage/utility' claim in the horizon is matched or beaten by a classical method (tensor-network / Pauli-propagation) within ~18 months of publication.

Blue-light-activated photolyase repairs UV-damaged DNA (cyclobutane pyrimidine dimers) through a cascade of ultrafast electron transfers using the same flavin (FAD) radical-pair chemistry that cryptochromes inherited and repurposed for candidate magnetoreception. Unlike the magnetic-compass claims, this repair chemistry is femtosecond-spectroscopy replicated across labs and organisms.

★ For the kids: Some sunlight-activated proteins fix broken DNA using the exact same 'radical pair' spark that their cousin proteins use as a maybe-compass — the repair job is proven; the compass job is still a maybe.

Technical: DNA photolyase splits UV-induced cyclobutane pyrimidine dimers via light-driven, multistep electron transfer from excited FADH- through a chain of aromatic residues to the lesion, generating short-lived radical-pair/radical intermediates resolved on the femtosecond-to-nanosecond timescale (Kao, Y.-T. et al., PNAS 2005; mechanism reviewed by Sancar, A., Chem. Rev. 2003). Photolyase and cryptochrome are structural paralogs; cryptochrome's radical-pair magnetosensitivity is widely thought to be an evolutionary exaptation of this repair chemistry. SKEPTIC NOTE: viable tier upheld — this is femtos…

Frozen prediction: Within the horizon, no published measurement overturns the femtosecond electron-transfer/radical-intermediate sequence established for photolyase DNA repair across at least two independently studied organisms.

Skeptic self-audit demotion (modeled -> open). The claim rests largely on one transgenic-fly rescue: hCRY2 restored light-dependent magnetic response in Cry-null Drosophila (Foley 2011). But the fly phenotype it rescues is itself now contested by a ~100,000-fly null (Bassetto 2023). A rescue of a contested phenotype cannot outrank the phenotype's own tier.

★ For the kids: A human eye-protein seemed to give fruit flies back a magnet sense in one experiment — but the fly magnet sense itself may not be real, so this drops down the ladder too.

Technical: Foley, Gegear & Reppert 2011 — hCRY2 rescues light-dependent magnetosensitivity in Cry-null flies. The assay's validity depends on Drosophila magnetosensitivity, which failed a high-powered replication. No in-vitro hCRY2 radical-pair MFE of Xu-et-al. CRY4 caliber exists.

Frozen prediction: No high-powered assay establishes robust hCRY2-mediated human magnetosensitivity within the horizon.

SKEPTIC DEMOTION (modeled -> open): static magnetic fields are reported to alter cryptochrome-dependent blue-light growth responses (hypocotyl elongation, anthocyanin accumulation) in Arabidopsis thaliana, consistent with a radical-pair mechanism operating in plant cryptochromes. But this rests on a SINGLE 2007 study whose own text concedes small effect sizes and thin independent replication — and roughly 19 years later, no independent blinded replication has been located. A single small-effect, unreplicated result does not clear the bar for 'modeled' (mechanistically well-supported or multip…

★ For the kids: Even plants might carry a tiny bit of the same 'quantum compass' chemistry — a nearby magnet can slightly change how a seedling grows toward blue light. But only one team has shown this, the effect was small, and nobody has double-checked it in almost 20 years, so we're calling it a maybe, not a modeled finding.

Technical: Ahmad, M., Galland, P., Ritz, T., Wiltschko, R., Wiltschko, W. (2007) 'Magnetic intensity affects cryptochrome-dependent responses in Arabidopsis thaliana,' Planta 225(3):615-624 — near-null vs ~500 uT static fields shift cry1/cry2-dependent hypocotyl growth and anthocyanin levels under blue light; effect absent in cry-deficient mutants, consistent with (not proof of) radical-pair involvement. Small effect sizes and the absence of any located independent replication in ~19 years keep this at open, not modeled, per skeptic audit.

Frozen prediction: Within the horizon, an independent lab replicates a cryptochrome-dependent Arabidopsis magnetic-field growth effect under blinded, randomized field conditions.

HELD at open, entered defensively. Magnetogenetics claims weak fields can ACTUATE neurons — field in, ion flux out — and the standing objection is energetic: the available magnetic energy per construct falls orders of magnitude short of gating a channel. The 2026 engineered-protein result is categorically different: field in, FLUORESCENCE out. Reading a spin state optically costs nothing the photon budget does not already supply; forcing a conductance change does. The energy objection is untouched, and citation-laundering here is pre-registered as a refusal.

★ For the kids: Two different things people mean by 'magnetic proteins'. One: a protein that CHANGES ITS GLOW near a magnet, so you can watch it. That one now works. Two: a protein a magnet can PUSH hard enough to switch a brain cell on. That still has a big problem — the pull is far too weak, like knocking over a house by blowing on it. The first working does not fix the second.

Technical: Magnetic energy per construct << kT at physiological field strengths and realistic moments. Abrahams et al., Nature 649:1172 (2026) is a spin-dependent OPTICAL readout — it changes recombination yield and hence fluorescence; it transduces no force and gates no conductance. Pre-registered refusal: any magnetogenetics paper citing MFP/ODMR results as mechanistic support without addressing the energy budget is scored as citation laundering, not evidence. The legitimate adjacency runs the other way — MFPs give sceptics a calibrated in-cell magnetometer for testing whether claimed actuator constru…

Frozen prediction: By 2029-12-31 no peer-reviewed paper resolves the magnetogenetics energy-budget objection with a MEASURED (not modelled) per-construct magnetic energy sufficient to gate an ion channel; and at least one paper cites engineered magneto-fluorescent-protein work as support for magnetogenetic actuation without addressing that budget.

VERIFIED: Legg, 'On the robustness of topological gap detection via transport', Nature Matters Arising (2026; doi:10.1038/s41586-026-10567-8) — transport data show disordered, gapless devices; parity signal consistent with a fine-tuned mesoscopic effect. Microsoft Reply (doi:10.1038/s41586-026-10568-7) holds the parity-readout claim does not depend on TGP. VERIFIED: 'Majorana 2' unveiled at Build, 2 Jun 2026, claiming ~20 s coherence (1,000x) on a corporate preprint with no X-measurement — a long-lived parity state, not a demonstrated qubit. Corporate preprint = dated slip-clock snapshot, not…

★ For the kids: A company says it built a super-stable qubit. Other physicists checked the raw data and say the chips look too messy. The company admitted a counting mistake but says it doesn't matter. Still unproven.

Technical: Legg, H., Nature (2026), doi:10.1038/s41586-026-10567-8 — TGP lacks consistent definitions of 'gap' and 'topological'; outcome sensitive to magnetic-field range, bias-voltage range, data resolution and cutter-voltage pair count, varying even between measurements of the same device. Prior arXiv comments (2502.19560, 2503.08944) are pointers. Majorana 2 announced at Build 2026 citing a preprint with >20 s coherence; critics maintain the device has not been shown to function as a qubit. RETIREMENT TRIGGER (frozen): if by 2027-06-30 no group free of Microsoft authorship and funding publishes a pe…

Frozen prediction: By 2027-12-31 no peer-reviewed reproduction of h/2e-periodic single-shot parity readout with topological attribution appears from a group with no Microsoft affiliation or funding (p=0.85).

Held at viable: this is a measured deployment fact from independent vantage points (arXiv:2606.16473 internet-wide measurement: 49.3% of domains with hybrid PQ key exchange; Cloudflare 2025/2026 telemetry; Let's Encrypt June 2026 MTC roadmap: staging late 2026, production 2027), not a physics claim. Operator telemetry is non-peer-reviewed and flagged. Mechanism: ML-DSA-44 ~2,420-byte signatures compound per-connection against initial congestion windows. NIST HQC backup KEM selected 11 March 2025; FIPS 203/204/205; EO 14412 ~2035 horizon.

★ For the kids: The internet needs two quantum-proof locks: one keeps secrets, one proves who you are. The first is on about half the internet; the second is on almost none because its signatures are too big.

Technical: Harvest-now-decrypt-later bites on confidentiality, the half that IS deploying — cuts against urgency framing.

Frozen prediction: By 2028-08-21 fewer than 5% of publicly-trusted TLS certificates in a major internet-scale scan use a PQ signature or MTC scheme: P=0.7.

Squeezed vacuum states of light are ALREADY deployed in Advanced LIGO to beat the photon shot-noise limit, directly increasing the detectors' astrophysical reach — quantum sensing literally extends astronomical range today.

★ For the kids: LIGO, the machine that 'hears' colliding black holes, uses a quantum trick with light to listen more quietly and catch fainter ripples in space — like quantum noise-cancelling headphones for the universe.

Technical: Tse et al., Phys. Rev. Lett. 123, 231107 (2019); frequency-dependent squeezing later extended detection range further in subsequent observing-run upgrades. Deployed and operationally validated across multiple observing runs — viable.

Frozen prediction: A next-generation gravitational-wave detector design (e.g. Einstein Telescope / Cosmic Explorer) formally adopts frequency-dependent squeezing as baseline within the horizon.

Small multi-node quantum networks — entangling remote solid-state qubits over real fiber, and telecom-band photonic quantum memories — have been demonstrated at small scale. Current demos are node-count-limited and short-range relative to a genuinely useful network.

★ For the kids: Scientists have linked a few quantum computers together with light, like a tiny practice version of a quantum internet — it works, but it's still just a few machines over a short distance.

Technical: Pompili et al., Science 372, 259-264 (2021); Knaut et al., Nature 629, 573-578 (2024). Open questions: metro-distance coherence scaling, routed (not linear-chain) networks.

Frozen prediction: A quantum-repeater or memory-node network demonstrates entanglement distribution across four or more nodes end-to-end within the horizon.

One scaling path links smaller quantum processors via photonic interconnects and runs algorithms distributed across them. An early two-module demonstration exists; whether this out-scales monolithic chip approaches is unresolved.

★ For the kids: Instead of building one giant quantum computer, some teams are trying to plug several smaller ones together with light-based cables, like connecting computers into a team instead of building one huge machine.

Technical: Main et al., 'Distributed quantum computing across an optical network link,' Nature (2025) — Oxford trapped-ion QCCD modules linked photonically. Open questions: fidelity/rate vs monolithic chips, winning qubit modality.

Frozen prediction: A distributed quantum computation is demonstrated scaling to three or more networked processing modules within the horizon.

Survival clock from Oct 2025 (Google Quantum AI, Nature 646, 65 qubits, OTOC(2)). VERIFIED: Bermejo, Villalonga, Ware, Vidal, Szasz, arXiv:2604.15427 (16 Apr 2026) — the abstract page lists no affiliations; Villalonga, Vidal, Ware and Szasz are Google Quantum AI-associated by the researcher's knowledge, so it is filed DEFENDER-RUN pending a PDF affiliation read. VERIFIED: Kshetrimayum, Jahromi, Singh, Orus, arXiv:2603.18825 (Mar 2026) — no affiliations on the abstract page; Orus is a Multiverse Computing co-founder, an interested party on the tensor-network side. Exact classical verification …

★ For the kids: Google says its chip did a calculation no normal computer can. So far the people checking hardest are mostly Google's own scientists, and the outside checkers sell the competing method.

Technical: Independent = no author with Google Quantum AI affiliation or funding; classical reproduction = OTOC(2) values at reported error bars for the 65-qubit deepest-cycle instance, or a demonstrated spoof of the verification statistic.

Frozen prediction: By 2027-04-30 no peer-reviewed independent classical reproduction of the 65-qubit OTOC(2) data exists (p=0.60) — graded symmetrically: a hit keeps the claim alive but unverified; a miss retires it.

VERIFIED: Khan et al., eNeuro 11(8) (2024; epothilone B delays isoflurane LORR in rats, Wiest lab); Huang et al., Neuropharmacology 287:110834 (Apr 2026; same direction in mice, same lab, Wiest senior); Li et al., BMC Anesthesiology 25:109 (2025; MGH, Martyn) — the only study without Wiest/Hameroff authorship — found epothilone D and vinblastine INCREASED isoflurane sensitivity (leftward LORR), paclitaxel marginal rightward, morphine rightward. The independent leg does not replicate; it complicates (different drugs, chronic >2-week regimen). Advocacy syntheses (Wiest, Neurosci Conscious 2025 …

★ For the kids: One lab found a drug that keeps rats awake longer under anesthesia and said it supports the quantum-mind idea. A different hospital lab tried similar drugs in mice and got the opposite result.

Technical: Microtubule drugs alter axonal transport, vesicle cycling and spine morphology — classical routes to potency shifts. A quantum-dampening picture predicts one sign; classical pharmacology allows either sign and regimen dependence.

Frozen prediction: By 2027-12-31 no peer-reviewed rodent study from a lab with no Wiest or Hameroff co-authorship reports that a brain-penetrant microtubule stabilizer DELAYS isoflurane LORR (p=0.70).

PROMOTION REFUSED, held at modeled. The hard part happened: the nuclear isomeric transition was laser-excited by independent groups in 2024 and its frequency tied to an optical atomic clock. That is replication-grade evidence for the TRANSITION. It is not evidence for a CLOCK. No thorium-229 device has been operated as a frequency standard, and its projected insensitivity to external fields — the entire point — remains projected. Worth stating out loud because the headline pressure here runs toward viable.

★ For the kids: Scientists finally managed to make an atom's tiny nucleus tick, which people chased for fifty years. But ticking is not the same as being a clock — nobody has actually used it to tell the time yet.

Technical: Tiedau et al., PRL 132, 182501 (2024) in Th:CaF2; independently Elwell et al., PRL 133, 013201 (2024). Frequency ratio against the Sr-87 lattice clock: Zhang et al., Nature 633, 63 (2024), ~1e-12 fractional uncertainty. Absent: continuous interrogation, a measured systematic uncertainty budget, and any demonstration of the predicted suppressed EM sensitivity. Comparison class: optical lattice clocks at 8e-19 (Aeppli et al., PRL 2024).

Frozen prediction: No thorium-229 nuclear clock reports a full systematic uncertainty budget below 1e-16 in a peer-reviewed journal by 2028-12-31; promotion to viable blocked until it does.

SKEPTIC DEMOTION modeled -> open. The magnesium-25 magnetic-isotope effect on ATP-producing enzymes (Buchachenko) is a physics-grounded but largely single-group body of work that has not consolidated into broad independent replication across enzyme systems. Real spin chemistry is plausible; the enzyme-kinetics claim is not replication-grade and stays open pending a second independent system.

★ For the kids: Some atoms carry a tiny inner magnet (nuclear spin). Swapping to the magnetic version may speed an enzyme up — an interesting clue, but shown mostly by one group so far.

Technical: Buchachenko et al. — Mg-25 (I=5/2) vs spinless Mg-24/26 on ATP synthesis (PNAS 2005 + follow-ups). Distinguishes hyperfine (spin) from classical mass/kinetic isotope effect; independent cross-system confirmation is thin.

Frozen prediction: A spin-dependent magnetic isotope effect is demonstrated in a second, independent enzyme system with a nonradioactive isotope pair within the horizon; absent that it stays open.

Held OPEN (speculative). Reports that lithium's behavioural effects and hypomagnetic-field brain effects depend on nuclear spin invite a radical-pair explanation, but behavioural isotope data are sparse and confounds (mass, kinetics) are not excluded. Modeling, not established mechanism.

★ For the kids: Early, uncertain hints that which 'version' of lithium changes animal behaviour, and that shielding animals from Earth's magnetism affects brain-cell growth. Might be quantum spin — still a maybe.

Technical: Radical-pair modeling of Li-7 vs Li-6 behavioural/circadian effects and hypomagnetic ROS-mediated neurogenesis (Zadeh-Haghighi & Simon, Sci. Rep. 2021; J. R. Soc. Interface 2022).

Frozen prediction: No replication-grade behavioural lithium-isotope effect with a spin-specific radical-pair signature is established within the horizon; a clean replication would promote to modeled.

FLAGGED FRONTIER, carried at open and never above: Fisher 2015 proposed P-31 nuclear spins in calcium-phosphate (Posner) clusters could store quantum information for cognitively-relevant times because nuclear spins decohere far slower than electron spins. This is spin CHEMISTRY, not demonstrated awareness; both key premises remain unverified.

★ For the kids: A bold idea says the brain might hide information in the spins of phosphorus atoms — interesting, but far from proven, and it does NOT mean the brain is a quantum mind.

Technical: P-31 (spin-1/2) in Ca9(PO4)6 clusters proposed to have second-to-day coherence; entangled Posner pairs could bias glutamate release. Motivating anecdote: lithium-isotope behavioral differences. Unverified: Posner coherence times in vivo, and any actual link to processing.

Frozen prediction: Advances from open ONLY if BOTH lithium-isotope behavioral effects reproduce dose-matched AND Posner nuclear coherence exceeds ~seconds in vitro; awareness stays unasserted regardless.

MERGED and HELD at viable. NV magnetometry, OPM-MEG, hyperpolarised NMR and quantum-limited photodetection are quantum INSTRUMENTS measuring classical biology — deployed, commercial, replication-grade. None of them is evidence that biology is quantum. A 2026 systematic evidence map independently arrives at the same three-way split and the same tiering, which raises confidence in the GRADING PROCEDURE and in nothing about the underlying physics.

★ For the kids: Two different sentences: 'we used a quantum tool to study a living thing' and 'the living thing is doing quantum things'. The first is normal and works great. The second is rare and hard to prove. Someone else drew their own map of the field and it looks a lot like ours — that doesn't make any single claim more true.

Technical: Three-direction map (Gassab, Adams, Goolam Hossen, Hassasfar, Pusuluk, Murugan, Kominis, Müstecaplıoğlu, Petruccione & Craddock, arXiv:2605.00205, 2026 — PREPRINT, corroborating only, no tier weight): quantum IN biology; quantum FOR biology, gated on beating strong classical baselines under realistic calibration, dose, throughput and uncertainty constraints; biology FOR quantum. Its stated tiering — constrained enzymatic H-tunnelling and radical-pair spin chemistry most mature, higher-visibility topics unresolved under physiological conditions — matches this corpus. USE RULE: it may not promo…

Frozen prediction: An NV-diamond or optically-pumped magnetometer resolves an in-vivo, field-correlated signal in candidate magnetoreceptive tissue within the horizon WITHOUT, by itself, promoting any quantum-IN-biology mechanism claim above its current tier.

Held MODELED. Encoding a logical qubit in a noise-biased oscillator 'cat' state suppresses the dominant bit-flip errors with far fewer parts. AWS Ocelot (2025) demonstrated concatenated bosonic encoding as a real second architecture — not yet a scaled logical machine, single platform.

★ For the kids: Instead of hundreds of qubits to fix mistakes, store the info in a wobbling 'cat' that mostly can't make the most common mistake in the first place.

Technical: Noise-biased cat qubits (bit-flip time exponential in photon number) + outer repetition code for residual phase flips; AWS Quantum, Nature 2025. Complements the transmon surface-code path.

Frozen prediction: A cat-qubit logical memory shows error suppression across >=2 concatenation/distance steps on a second independent platform within the horizon.

Held MODELED. A neural-network decoder (AlphaQubit, 2024) beat hand-built decoders on real surface-code data, but real-time throughput at scale remains the open engineering wall.

★ For the kids: A quantum computer needs a fast helper to spot its mistakes; scientists trained an AI to be better at it than the old rulebooks.

Technical: Recurrent-transformer decoder trained on Sycamore data outperformed MWPM/correlated matching at d=3,5 (Bausch et al., Nature 2024). Latency/scalability to real-time large-distance decoding unproven.

Frozen prediction: An ML decoder is demonstrated inside a real-time feedback loop on a distance>=5 code within the horizon.

DEMOTION UPHELD at open, and the scoping is the important part. The on-hardware protocol's security REDUCES to the classical hardness of random-circuit sampling — the exact claim this ledger holds at open — so the derived product was being rated above its own foundation. The demotion is scoped: Bell-based device-independent randomness rests on loophole-free Bell tests (viable) and is genuinely stronger; it is the RCS-certified variant that cannot outrank RCS, and the two must be tracked separately rather than merged under one confident headline.

★ For the kids: A quantum computer can make numbers nobody could have guessed and prove it — but the proof leans on the belief that ordinary computers can't copy the machine. We're not sure of that belief yet, so we can't be surer of the proof than of the belief.

Technical: Aaronson & Hung, STOC 2023: certified randomness from quantum supremacy, security conditional on RCS spoofing hardness with classical cross-entropy spot-checking. Liu et al., Nature (2025): trapped-ion demonstration. If RCS hardness erodes — as it has for several sampling claims — the certification degrades with it. Bierhorst et al., Nature 556, 223 (2018): Bell-based certification does not share the dependency.

Frozen prediction: RCS-certified randomness is not promoted above open before the underlying sampling-hardness claim survives 24 months with no classical-spoofing result; Bell-based certification tracked separately.

Held VIABLE. Entangling atoms in an optical clock (spin squeezing) reduces quantum projection noise below the standard quantum limit; demonstrated in optical-lattice/tweezer clocks (2023-2024) with independent metrological gain — replication-grade quantum-enhanced sensing. Distinct from the base optical-lattice-clock component: this is the entanglement-enhancement axis specifically.

★ For the kids: Normally each atom in a clock adds jitter; gently entangle them and the jitter partly cancels, so the clock steadies faster.

Technical: Rydberg one-axis-twisting squeezing (Eckner et al., Nature 2023); two-ensemble comparison at 1e-17 (Robinson et al., Nat. Phys. 2024). Near-term question: gain at the systematics-limited frontier of the best clocks.

Frozen prediction: A spin-squeezed clock reports a metrologically useful gain while operating at or below 1e-18 fractional uncertainty within the horizon.

Held VIABLE. Simultaneous closure of the detection + locality loopholes (Hensen et al. 2015; Giustina/Shalm et al. 2015; 2022 Nobel) makes entanglement/nonlocality a replicated fact, not an interpretation — the foundation under certified randomness and QKD security theorems.

★ For the kids: Scientists proved, with no cheating room left, that two far-apart quantum particles really are spookily linked.

Technical: NV electron spins 1.3 km apart (Hensen et al., Nature 2015) and high-efficiency photon experiments (Giustina; Shalm, PRL 2015); CHSH violation with space-like separation. Only the practically untestable freedom-of-choice/superdeterminism loophole remains.

Frozen prediction: No credible local-hidden-variable reinterpretation of a loophole-free Bell violation gains traction within the horizon.

Demotion upheld. The pattern (Sycamore matched in ~3 years: Pan, Chen & Zhang PRL 129, 090502, 2022; IBM utility matched in weeks: Tindall et al. PRX Quantum 5, 010308, 2024, Begusic et al. Sci. Adv. 2024; D-Wave annealing partially matched in ~14 months: Science 2026) stays externally anchored under classical-refutation-pattern. The three-channel laundering taxonomy is engine-authored and uncited; open until an external source names and counts at least one channel.

★ For the kids: The pattern that quantum 'firsts' get caught up by normal computers is real and shown by other scientists. Our tidy list of the tricks involved is our own idea, so we label it as a guess.

Technical: Clock arithmetic: survival_time counted only while at least one zero-overlap party is known to be attempting refutation; with no independent attempt the claim is marked UNTESTED, never SURVIVING — the distinction channels 2 and 3 erase. Overlap computed at institution level for corporate labs, senior-author level for academic labs. Withdrawn preprints scored as non-existent. Currently UNTESTED (not surviving): Quantum Echoes OTOC advantage; the epothilone-B anaesthetic-latency effect; myelin biphoton generation; human single-photon perception (10 years); KcsA filter coherence (16 years); MagL…

Frozen prediction: By 2027-08-21 an external peer-reviewed or formal-audit source names and counts at least one advantage-claim laundering channel: P=0.4. If none, stays open.

HELD at open, but the reasoning has changed and that matters. The field leaned toward the Trp-tetrad partly on the physical argument that superoxide is too hyperfine-poor and too fast-relaxing to yield Earth-strength effects. A 2024 simulation shows strongly asymmetric recombination puts such a pair in a quantum-Zeno regime that restores sensitivity. Nothing has been measured. Net: neither candidate can be excluded on theory, neither has in-protein experimental discrimination at Earth-strength fields, and Trp-tetrad should stop being treated as the default.

★ For the kids: The bird-compass idea needs TWO partners: one is known, the other is a mystery with two suspects. People mostly ruled out the second suspect because it seemed physically incapable — new calculations found a loophole. So both suspects are back in the running and nobody has proof either way. Learning something that makes you LESS sure is still learning.

Technical: Denton, Smith, Xu, Pugsley, Toghill & Kattnig, Nat Commun 15:10823 (2024) removes the hyperfine-poverty exclusion for FADH−/O2− under asymmetric recombination, at the cost of requiring near-total immobilisation of the superoxide — in tension with the escape/signalling channel biology needs. Discriminating test: isotope substitution or scavenger manipulation on a protein-embedded pair with direct spin readout at ≤100 µT. NOTE: this component now consumes a theory-only rescue, so anything consuming THIS component inherits open at best.

Frozen prediction: By 2029-12-31 no experiment discriminates the Trp-tetrad from the superoxide partner in a protein-embedded cryptochrome radical pair at Earth-strength fields.

SKEPTIC DEMOTION (modeled -> open). Ikeya & Woodward 2021 report single living HeLa-cell autofluorescence changing a few percent under weak fields, consistent with flavin radical-pair chemistry. But this is a SINGLE-LAB, unreplicated, few-percent effect whose in-cell radical identity (superoxide vs Trp) is itself the marginal branch (Player & Hore). 'Modeled' overreaches for an unreplicated, surprising in-vivo MFE — held at open pending independent replication and the RF-resonance signature.

★ For the kids: One lab saw a magnet slightly change the faint glow of ordinary cells — an interesting hint, but nobody else has repeated it yet, so it stays a maybe.

Technical: Flavin photoexcitation generates [FADH•/superoxide or Trp•] pairs; field-dependent singlet-triplet yield alters fluorescent-vs-dark branching (~3-4% modulation). First claimed optical radical-pair MFE in a non-specialized mammalian cell, but no second lab and no Zeeman-matched RF suppression control yet.

Frozen prediction: Promotes to modeled only on independent replication PLUS a Zeeman-matched RF-resonance suppression of the effect by 2027.

Russo et al. (2026) widen the iron signal without localizing it or testing polarity sensitivity. Consistent with an ancient iron-handling trait, dietary iron, or a distributed receptor. Held at open.

★ For the kids: Lots of bees carry magnetic specks, but the specific honeybee compass organ has not been pinned down.

Technical: Bridge to the bee crew: watch-item, not a navigation input.

Frozen prediction: By 2028-03-05 no peer-reviewed single-domain magnetite localization with innervation is shown in a honeybee tissue using element-specific imaging (p=0.65).

Shielding away the geomagnetic field perturbs biology — ROS balance, development, reported adult neurogenesis — consistent with but not proof of a radical-pair mechanism. Carried as MECHANISM biology (spin chemistry / redox), explicitly NOT awareness.

★ For the kids: Take Earth's magnetic field away and living cells change a little — hinting life quietly 'feels' the field through quantum-ish chemistry.

Technical: Hypomagnetic effects on ROS and cellular function are analysed within the radical-pair framework (Zadeh-Haghighi & Simon 2022); reported in-vivo rodent neurogenesis effects are not pinned to a spin-correlated radical pair vs generic redox signalling. Open: spin-correlated radical pair vs downstream redox, and cross-lab/species replication.

Frozen prediction: No hypomagnetic in-vivo mammalian effect is shown to REQUIRE a quantum radical-pair mechanism (vs generic ROS chemistry) under a pre-registered RF/isotope-controlled test within the horizon.

Held at modeled — not promoted, and the proposers' 'strengthened' language is trimmed. The turtle map sense now has what avian magnetoreception has never obtained: causal manipulations under an assay that isolates map from compass. A magnetite-flipping pulse disrupts it; a weak magnet on the head disrupts it while the same magnet on body or tail does not. The tier holds rather than rises because it remains one laboratory, one species, receptor unidentified, and the pulse diagnostic is suggestive rather than probative. The honest reading for this lens is uncomfortable: two mechanisms in one an…

★ For the kids: Baby sea turtles know where they are in the ocean using magnetism. A strong magnetic zap breaks it — what you'd expect if tiny iron crystals do the sensing — and a magnet on the head breaks it while one on the tail doesn't.

Technical: Mackiewicz et al., J Exp Biol 228(22):jeb251243 (2025), doi:10.1242/jeb.251243; Lim et al., J Exp Biol 229(10):jeb252113 (2026), doi:10.1242/jeb.252113 — both verified. The modeled tier survives the single-lab rule applied elsewhere in this pass because the turtle magnetic-map behaviour itself has a multi-decade published record and the magnetic-map concept has independent support in other taxa; what is single-lab is the localization, which is carried separately at open. Navigation context this cycle: Hays GC et al., Sci Adv (2026), doi:10.1126/sciadv.aed8113, mid-ocean reorientation in long-…

Frozen prediction: Carried by localized-magnet-receptor-localization. Additional: P=0.15 that by 2029-12-31 a light-dependence test on the turtle MAP sense (map disruption under monochromatic long-wavelength vs full-spectrum light) is published; if run and the map proves light-independent, that is the cleanest available separation of the two senses in one animal.

SELF-AUDIT OF THE SKEPTIC'S OWN LOAD-BEARING NUMBER, held at modeled. Skeptics cite Tegmark's ~10^-13 s microtubule decoherence estimate as if it were a measurement. It is not, and neither is any number published against it: the spread across the literature is roughly sixteen orders of magnitude, with ZERO in-vivo measurements. The GENERAL physics — a warm, wet, strongly coupled environment destroys phase coherence fast — is solid and is why the burden sits with the advocate. The SPECIFIC microtubule number is unconstrained by experiment, and a skeptic quoting 10^-13 s as fact commits the sam…

★ For the kids: Warm wet things wreck delicate quantum effects fast — real physics, and the main reason to doubt quantum-brain ideas. But ask 'how fast, exactly, inside a real brain cell?' and every answer is somebody's calculation, and the answers disagree enormously. Nobody has measured it. A careful doubter has to admit that too.

Technical: Documented spread, all theoretical: Tegmark ~10^-13 s; 10^-16 to 10^-8 s reported in Ma & Wang, Front Psychol (2026), doi:10.3389/fpsyg.2026.1730965, which also floats 10^-9 to 10^-12 s as GHz–THz-compatible; and 10–100 µs claimed by Sergi A. et al., Front Hum Neurosci 19:1630906 (2025) — a paper explicitly authored to be supportive-reformulative of Orch OR, which is disclosed because a model written to rescue a theory is not an independent constraint on it. Estimates trend monotonically upward over 25 years with not one revision driven by a new measurement. STANDING RULE ADOPTED: no maturity…

Frozen prediction: By 2028-01-01 no peer-reviewed direct measurement of a coherence or dephasing time for any degree of freedom in a microtubule within a living neuron is published.

Entangling spatially separated probes lets a sensor NETWORK beat the standard quantum limit on a GLOBAL parameter (a field average or gradient). Genuine below-SQL gains shown in the lab, but for global parameters and under idealized loss — modeled, not yet fielded.

★ For the kids: If lots of tiny quantum sensors share a spooky link, together they can measure a shared thing more precisely than any one alone.

Technical: CV entangled networks improve estimation of a distributed parameter below the SQL (Guo et al., Nat. Phys. 2020); mode-entangled spin-squeezed ensembles show a distributed gain (Malia et al., Nature 2022). Advantage is for GLOBAL/joint parameters; per-node local estimation is not automatically enhanced. Open: survival under realistic per-node decoherence, and which real task beats a single squeezed probe.

Frozen prediction: A below-SQL distributed-sensing advantage is demonstrated on a second physical platform, for a genuinely multi-node global parameter, within the horizon.

Highly-excited Rydberg atoms in a vapour cell turn an RF electric field into an optical readout via EIT — an atom-based, SI-traceable, self-calibrating RF receiver/field sensor. Real and traceable, but must still prove a sensitivity-bandwidth win over classical antennas.

★ For the kids: Puff atoms up until they're huge and they act like tiny radio antennas you read out with a laser — a self-calibrating quantum radio.

Technical: EIT / Autler-Townes splitting in a Rydberg ladder gives absolute E-field measurement (Sedlacek et al., Nat. Phys. 2012); since demonstrated as SI-traceable field probes. Open: a genuine sensitivity-bandwidth advantage over antennas, not just traceable calibration.

Frozen prediction: A Rydberg-atom electrometer is adopted beyond its originating lab as an SI-traceable RF field standard within the horizon.

Freely-falling cold atoms in a matter-wave interferometer measure gravity and inertial forces at quantum-limited precision — replicated, commercialized gravimeters plus lab-scale equivalence-principle tests. Quantum sensing that already works in the field.

★ For the kids: Drop super-cold atoms and use their quantum wave nature as a ruler to weigh gravity itself.

Technical: Light-pulse atom interferometry; transportable gravimeters at uGal sensitivity; equivalence-principle test to 10^-12 (Asenbaum et al., PRL 2020); source-mass gravitational Aharonov-Bohm phase observed (Overstreet et al., Science 2022). Bridges to geodesy, dark-sector searches (MAGIS/AION), long-baseline GW concepts.

Frozen prediction: A transportable cold-atom gravimeter is cross-validated against a superconducting gravimeter at the uGal level in a field/geodesy campaign within the horizon.

A structural skeptic finding: flagship 'quantum advantage/utility' demos have repeatedly been matched or beaten by improved classical algorithms within ~1-2 years. Treat every fresh advantage claim as classically-explicable until a durable classical null survives. The hype axis's central discipline.

★ For the kids: Every time a quantum computer 'wins,' clever people often find a way for a normal computer to do the same thing soon after. So we wait before believing 'quantum won.'

Technical: Sycamore RCS tightened by tensor networks (Pan, Chen & Zhang, PRL 2022); IBM 127-qubit 'utility' Ising (Kim et al., Nature 2023) reproduced classically (Tindall et al., PRX Quantum 2024; Begusic et al., Sci. Adv. 2024); GBS spoofed (Oh et al., Nat. Phys. 2024). XEB has known loopholes (Gao et al., PRX Quantum 2024).

Frozen prediction: At least one 2023-2025 'quantum advantage/utility' claim is matched or beaten by an improved classical algorithm within the horizon.

Held at open. The '20x in six years' compression is real as arithmetic and unbuilt as hardware: Gidney's 2025 preprint puts RSA-2048 under a week with fewer than a million noisy qubits, against 20 million in Gidney & Ekerå (2021). The new estimate stacks three techniques each carried at modeled — approximate residue arithmetic, yoked surface codes, and magic-state cultivation — so it inherits the weakest tier of its ingredients. The correct reading, and the one the corpus should repeat: the algorithmic clock is the moving part; the hardware clock has not moved by a comparable factor, and the …

★ For the kids: Someone worked out that a code-breaking quantum computer might need twenty times fewer parts than we thought. Nobody has built even a hundredth of it, and the recipe assumes parts better than any we can make today.

Technical: Gidney C, 'How to factor 2048 bit RSA integers with less than a million noisy qubits', arXiv:2505.15917 (21 May 2025, Google Quantum AI) — verified against the arXiv record; not peer-reviewed as of this pass. Assumptions that decide everything: p_phys=10^-3, surface-code cycle 1 µs, reaction time 10 µs, 2D-local superconducting layout, ~1400 logical qubits with yoked storage. Sensitivity: at p_phys=5×10^-3 the estimate does not close; at 10^-4 it drops further. Ingredient attribution per Gidney's own summary: approximate residue arithmetic (Chevignard, Fouque & Schrottenloher 2024), yoked sur…

Frozen prediction: By 2028-06-30, P=0.80 that no group demonstrates a fault-tolerant modular multiplication of a ≥32-bit number on logical qubits with Λ>1. If one does, promote to modeled and re-open the harvest-now-decrypt-later entry.

HELD at modeled; the over-reach is in the NAMING, not the tier. The experiments are strong and independently reproduced on two hardware modalities, which is why this is not demoted. But these are digitally prepared, measurement-and-postselection-based realisations of anyonic exchange statistics, with no energy gap and no topological protection. Press framing as 'topological qubits achieved' conflates simulating exchange statistics with obtaining protected logic. The physics earns modeled; the headline does not earn more.

★ For the kids: Scientists made particles that remember being swapped around — genuinely cool, and done twice by different teams. But that's a re-enactment on a normal quantum chip, not a new kind of super-stable quantum memory.

Technical: Andersen et al., Nature 618, 264 (2023) (superconducting); Iqbal et al., Nature 626, 505 (2024) (trapped ion). Both are digital state preparations with active error handling, not equilibrium topological phases; protection would require a gapped Hamiltonian and a demonstrated below-threshold logical gate. Distinct from topological-qubits-majorana (open).

Frozen prediction: Braiding of ERROR-CORRECTED logical non-Abelian anyons with a demonstrated protected gate is not achieved by 2029-12-31; further demonstrations remain digital preparations.

Arrays of hundreds of individually trapped neutral atoms excited to Rydberg states form programmable many-body quantum simulators, reproduced across many groups. Robust platform; only specific 'beyond-classical' claims for particular instances remain the contested edge.

★ For the kids: Scientists line up hundreds of single atoms with lasers and puff them up so they push on each other, building a tiny tunable model of hard physics.

Technical: 51-atom quench dynamics (Bernien et al., Nature 2017); 256-atom programmable phases (Ebadi et al., Nature 2021); 2D antiferromagnets (Scholl et al., Nature 2021). Multiply reproduced.

Frozen prediction: An independent group reproduces a Rydberg-simulated quantum phase that resists exact classical simulation, cross-validated within the horizon.

Programmable photonic processors performed GBS tasks claimed classically intractable, a photonic route to 'quantum advantage'. Like other advantage claims, several instances have been narrowed or matched by improved classical algorithms.

★ For the kids: Machines made of light shuffle photons in a way a normal computer supposedly can't copy, but scientists keep finding clever tricks to copy it anyway.

Technical: Jiuzhang (Zhong et al., Science 2020) and Borealis (Madsen et al., Nature 2022) reported GBS advantage. Subsequent classical spoofing/approximation (Oh et al., Nat. Phys. 2024) has contested several claims; advantage is instance-specific and unsettled.

Frozen prediction: A published classical algorithm matches or spoofs a headline photonic boson-sampling advantage claim within the horizon.

Demotion upheld. The oscillations (Nature 2021, s41586-021-04257-w) are undisputed; the MBL eigenstate-order claim is contested by Khemani, Moessner & Sondhi (arXiv:2109.00551, preprint) and treated as unsettled by Zaletel et al. Rev. Mod. Phys. 95:031001 (2023). Prethermal DTC observations (PRX 13:041016, 2023) are not in dispute and not demoted.

★ For the kids: Some quantum computers made a pattern that repeats forever and people called it a new state of matter. Others say the wobble is real but ordinary, and experiments cannot yet tell.

Technical: Generic initial states decay rapidly; long-lived oscillations from polarised states follow from static transverse-field Ising physics.

Frozen prediction: By 2029-08-21 no MBL-DTC order demonstrated on a digital processor from generic initial states beyond the prethermal timescale: P=0.7. If shown, restore to modeled.

Quantum techniques — superconducting-qubit single-photon counting and squeezed-vacuum states — push microwave-cavity axion searches below the standard quantum limit. Emblem of quantum-enhanced fundamental-physics sensing; no detection yet, payoff is faster scan rate and deeper exclusion.

★ For the kids: To hunt invisible dark matter, scientists use quantum tricks to listen for signals quieter than the usual background noise.

Technical: Qubit-based photon counting evades the SQL for axion detection (Dixit et al., PRL 2021); HAYSTAC squeezed-state readout scans faster than the SQL (Backes et al., Nature 2021).

Frozen prediction: A quantum-enhanced (photon-counting or squeezed) axion search publishes a new sub-SQL exclusion limit within the horizon.

CONFIRMED viable. The only magnetic sense with a fully identified receptor: membrane-bound magnetite/greigite chains that passively torque the cell along field lines. Genes, biogenesis and quantitative physics are all characterised across independent laboratories over fifty years, and none of it is quantum. Carried as the field's benchmark for what 'solved' looks like and as the standing rebuttal to the assumption that a magnetic sense implies spin chemistry.

★ For the kids: Some bacteria grow a tiny row of iron magnets inside themselves and line up with Earth like a compass needle. Scientists know exactly how it works — and it isn't quantum at all.

Technical: Blakemore, Science 190, 377 (1975); magnetosome-island mam/mms genetics and biogenesis reviewed in Uebe & Schuler, Nat. Rev. Microbiol. 14, 621 (2016). A ~20-crystal chain's magnetic moment exceeds kT at 50 uT, so alignment is passive magnetostatic torque — no transduction, no receptor cell, no radical pair.

Frozen prediction: No magnetosome-based magnetotaxis is shown to require a radical-pair or spin-chemical step by 2030-12-31.

CONFIRMED viable — and it is the correct anchor for what 'viable' means in this corpus. Electrons hop 10–20 Å between redox cofactors by quantum tunnelling, with rates falling exponentially in distance exactly as tunnelling predicts. Replicated across four decades and many independent groups, quantitative over ~12 orders of magnitude in rate, with no serious classical alternative. Notably absent from most 'quantum biology' marketing, which is itself diagnostic of how the field selects its headlines.

★ For the kids: Inside every cell, electrons cross gaps they shouldn't be able to cross — they tunnel, like walking through a wall. That's how your body gets energy from food, and it is the best-proved quantum trick in all of biology.

Technical: Moser, Keske, Warncke, Farid & Dutton, Nature 355, 796 (1992): log k falls ~1.2 decades per Å edge-to-edge across natural redox chains. Gray & Winkler, PNAS 102, 3534 (2005): pathway and coupling mapping in Ru-modified metalloproteins, beta ~1.1/Å. Nonadiabatic Marcus theory fits throughout. Distinct from enzymatic hydrogen (nuclear) tunnelling, carried separately.

Frozen prediction: No mainstream challenge to distance-dependent electron tunnelling as the mechanism of biological long-range electron transfer appears by 2030-12-31.

Two candidate receptors were proposed in Science within six months (type II canal hair cells; ferritin-loaded liver macrophages). Neither has an independent replication or causal-with-rescue result, so the negative claim stands — but with explicit retirement conditions instead of a diffuse 'nobody has found it'.

★ For the kids: For decades nobody could find the body part that feels magnetism. Now there are two candidates, so this claim finally has a way to be proven wrong.

Technical: Retire when either candidate gets an independent replication with causal manipulation and behavioural rescue; demote to modeled on an independent journal replication without rescue.

Frozen prediction: By 2028-09-05 neither candidate receptor has an independent peer-reviewed replication with causal manipulation (p=0.6).

HELD at modeled. The compass has a brain address: Cluster N is active only during night-time magnetic orientation, lesioning it abolishes the inclination compass while sun and star compasses survive, and ophthalmic trigeminal section does not impair the compass. That dissociation is the strongest anatomical evidence the compass is light-dependent and eye-based — but it localises a pathway without naming the sensor molecule, so it cannot carry the domain above modeled.

★ For the kids: Birds have a patch of brain that switches on only when they navigate by magnetism at night. Switch it off and they lose their compass — but they can still steer by the stars.

Technical: Mouritsen et al., PNAS 102, 8339 (2005) (IEG mapping, garden warblers and robins); Zapka et al., Nature 461, 1274 (2009) (Cluster N lesion vs V1 section); thalamofugal input, Heyers et al., PLoS ONE 2, e937 (2007). Necessary-but-not-sufficient for any cryptochrome claim.

Frozen prediction: Cluster N lesion effects replicate in a third night-migratory songbird species, and no study shows trigeminal section abolishing the INCLINATION compass, by 2029-12-31.

Demotion accepted. Nordmann et al. (Science 2025/26) found vestibular and mesopallial activation with no trigeminal signal in an unbiased screen; the beak receptor was retired in 2012 with no replacement; Lisowski et al. (Science 2026) route the iron candidate through the vagus. Whole-nerve ablation behaviour (Kishkinev 2013; Pakhomov 2018) is retained but is non-specific. A map sense may be real; the TRIGEMINAL attribution has lost both anatomical legs.

★ For the kids: Scientists thought the bird's magnetic map ran through a face nerve. A whole-brain search did not find that nerve lighting up, so it is back to a maybe.

Technical: Reinstate to modeled only with a magnetically responsive trigeminal afferent recorded alongside an identified receptor cell, or a demonstration that the pigeon screen's null reflects stimulus choice.

Frozen prediction: By 2028-09-05 no peer-reviewed electrophysiological recording of a magnetically responsive trigeminal afferent with an identified receptor cell appears (p=0.8).

HELD at modeled. Induction via ampullary electroreceptors in a conductive marine medium is the physically sufficient and behaviourally supported account in sharks and rays — replicated conditioned magnetic discrimination, microvolt-per-metre sensitivity. The 2025 pigeon result widens the mechanism's candidate SCOPE to a terrestrial animal using endolymph as the conductor, but that avian instance is contested and filed separately at open, and must NOT be used to lift this component.

★ For the kids: Sharks find things by feeling tiny electric tingles, and moving through a magnetic field makes exactly that kind of tingle. Everyone thought only sea animals could do this because you need salty water. Now it looks like a pigeon might do it inside the salty fluid of its own ear — but that bit is still an argument.

Technical: Elasmobranch case: ampullae of Lorenzini physiology plus conditioned discrimination in stingrays and rays; induction is sufficient in seawater. Scope extension: Nordmann et al., Science (2025) scRNA-seq of pigeon cristae. Constraint: Kattnig, arXiv:2602.23485 (2026). Tier discipline enforced: marine instance modeled on its own evidence; avian instance open, separately.

Frozen prediction: By 2030-12-31 no peer-reviewed study measures an induced electrical potential in an intact avian semicircular canal under a geomagnetic-strength field change of the magnitude the Nordmann et al. model requires.

HELD at open. A radical-pair compass needs its receptors held in a common orientation or the directional signal averages away. Two 2025–2026 developments moved this in OPPOSITE directions: robin CRY4a was shown to bind lipid bilayers in an ordered, non-random helical orientation; and the hybrid magnetite proposal claims the requirement can be dissolved rather than met. Neither is an in-vivo measurement.

★ For the kids: For a bird's quantum compass to work, all the tiny sensor proteins have to point roughly the same way, or their answers cancel out. New work shows the protein really does line up neatly when it sticks to a fatty layer in a test tube. Whether it lines up like that inside a living bird's eye is still unknown.

Technical: FOR: Hochmair, Wong, Solov'yov, Winklhofer, Xu et al., ACS Chem Biol 20(3):592 (2025) — PM-IRRAS resolves helix conformation, motional freedom and average orientation of membrane-associated ErCry4a. AGAINST THE REQUIREMENT: Hore, PNAS 123:e2524093123 (2026). No promotion because (a) synthetic supported bilayers, not photoreceptor membranes; (b) same consortium; (c) ordering relative to a membrane is not ordering relative to the bird's HEAD, which is what a compass requires. Pre-registered promotion condition: an in-situ orientational measurement of CRY4a in intact avian retinal tissue, from a…

Frozen prediction: By 2030-12-31 no peer-reviewed study reports an in-situ measurement of CRY4a orientational order within intact avian retinal tissue.

HELD at open. The sharpest quantitative objection to the avian compass: computed electron-spin relaxation in cryptochrome is fast enough to erase the field effect before it can be read out. The proposed rescue — a third, scavenger radical that amplifies sensitivity and buys relaxation resilience — is elegant and entirely theoretical. A mechanism that needs an unverified rescue to survive its own numbers has not earned 'plausible'.

★ For the kids: The bird's quantum compass may blur too fast to be read. One clever fix says a third helper molecule sharpens it up — but that fix has only ever been done on paper.

Technical: Kattnig, Solov'yov & Hore, Phys. Chem. Chem. Phys. 18, 12443 (2016): MD-informed relaxation rates in tension with the microsecond coherence the compass needs. Kattnig & Hore, Sci. Rep. 7, 11640 (2017): a scavenging third radical amplifies sensitivity. No experiment has identified such a scavenger or measured the amplification in a native receptor.

Frozen prediction: No experimental identification of a scavenger radical amplifying a cryptochrome magnetic-field effect in a native receptor by 2029-12-31.

HELD at modeled. The avian inclination compass works only within a narrow band of field intensity around ambient, yet birds re-tune to anomalous intensities after a few hours' exposure. This 'functional window plus acclimation' is a hard, long-standing behavioural constraint that any candidate receptor model must reproduce quantitatively — and almost no published spin-dynamics model is ever tested against it. Cheapest available falsifier in the domain; capped at modeled because the behavioural literature leans heavily on one research lineage.

★ For the kids: A bird's compass only works when the magnetism is close to the strength it's used to — but give it a few hours and it can learn a new strength.

Technical: Wiltschko & Wiltschko, Science 176, 62 (1972); Winklhofer et al., Proc. R. Soc. B 280, 20130853 (2013) (re-tuning to anomalously low intensities). A radical-pair receptor model must reproduce both window width and plasticity; most Cry4 spin-dynamics simulations report singlet-yield anisotropy only.

Frozen prediction: No published Cry4 spin-dynamics model reproduces BOTH the measured functional-window width and the acclimation plasticity by 2029-12-31.

VERIFIED preprint: Salvia et al., bioRxiv 10.64898/2026.06.30.735402. In wild-type AsLOV2 the FMN–cysteine radical-pair spin state modulates photoadduct formation via g-factor asymmetry above ~100 mT. Genuine spin-to-signalling link, but in a LOV domain, at 2,000–26,000x geomagnetic, preprint tier.

★ For the kids: In a plant light sensor the spin of two electrons decides whether a bond forms — but only in magnets thousands of times stronger than Earth's.

Technical: Delta-g mechanisms are field-proportional and irrelevant below mT; the compass regime is hyperfine-dominated.

Frozen prediction: By 2027-09-05 no journal paper demonstrates a spin-state-dependent signalling output in any cryptochrome at fields below 1 mT (p=0.85).

DEMOTED modeled → open. What replicates is CISS in DEVICE geometries — mAFM, Hall, photoemission junctions using chiral molecules as conductors. What this component claims is CISS operating inside a functioning protein's electron-transfer chain, and that has never been measured even once. Compounding it, no theory reproduces the measured magnitude and at least five mutually incompatible mechanisms were proposed in an eight-week window in 2026, so every biological CISS model inserts the polarisation as a tuned free parameter. The corpus already caps unobserved-instance claims at open (see cell…

★ For the kids: Some molecules are shaped like corkscrews, and electrons spiralling through them come out spinning mostly one way. That's real in test-tube devices. But nobody has caught it happening inside a working bird protein, and five different teams give five different reasons why the trick works at all.

Technical: Spin polarisation of tens of percent in junction geometries; ab-initio spin-orbit coupling in light-element organics undershoots by 1–3 orders of magnitude. Competing 2026 accounts: arXiv:2607.06858 (exchange), 2607.12720 (geometric SOC), 2607.00668 (spinterface), 2606.17823 (tunnelling amplification). Magnetoreception entry point is a singlet-triplet basis rotation at initialisation (Tiwari & Poonia, Phys. Rev. E 106, 064409, 2022); Gwynn, Adams & Petruccione, arXiv:2605.22130 (2026) show the enhancement is conditional and can REVERSE SIGN. Biological CISS claims should be read as existence …

Frozen prediction: No peer-reviewed paper reports a direct measurement of spin polarisation of the FAD-Trp electron-transfer chain inside an intact cryptochrome protein by 2029-12-31 (synthetic chiral bridges and simulations do not count).

DEMOTED modeled → open — while endorsing the FRAME. Replacing 'is the compass quantum?' with 'how close does it get to the physical bound on field sensing per unit volume per unit time?' is the best reframing this crew has been offered, because it converts a metaphysical argument into a number an experiment can miss. But the tier grades evidence, not elegance: this is a theory-only result from essentially one group, with no in-vivo or intact-tissue sensitivity measurement of any animal magnetoreceptor at the required precision. The corpus holds theory-only claims at open (see spin-relaxation-…

★ For the kids: Physics says there's a best-possible compass nobody can beat. So instead of arguing whether a bird's compass is 'quantum', ask how close it gets to the best possible — that's a question with a NUMBER. Great question. Nobody has measured the number in a live animal yet.

Technical: Kominis & Gkoudinakis, PRX Life 3, 013004 (2025), apply E_R = δB²·V·τ/(2μ₀) ≥ ħ to radical-pair magnetometers. Smith et al., Phys. Rev. Applied 25, 024074 (2026): interradical motion pushes attainable precision toward the bound. Optimality bounds: Smith et al., Quantum Sci. Technol. 9, 035041 (2024). Missing: any loss budget in the LIGO style enumerating every decoherence and readout-inefficiency channel with a number.

Frozen prediction: No publication reports an in-vivo or intact-tissue magnetic-sensitivity measurement placing a biological magnetoreceptor within a factor of 10 of the energy-resolution limit by 2028-12-31.

VERIFIED: Abrahams, Štuhec, ... Timmel, Miller, Ingaramo, York, Tetienne, Steel, Nature 649:1172-1179 (21 Jan 2026) — MagLOV ODMR in living bacteria at single-cell SNR. VERIFIED: Burd, Bagheri, Condon, Ingaramo, ... York, ... Boxer, Kasevich, Nature (Mar 2026; doi:10.1038/s41586-026-10282-4) — RF-resonant control of RFP–flavin radical pairs in transgenic C. elegans. VERIFIED: Maria Ingaramo and Andrew G. York (Calico) are on BOTH author lists, so the two papers are not independent lineages. Two proteins, two hosts, two lead groups earns plausibility (modeled); replication-grade is refused. Th…

★ For the kids: Two teams can now read electron spins inside living cells by watching them glow under radio waves. Two people worked on both papers, so it is not quite two separate checks.

Technical: RF at the electron Zeeman frequency alters singlet-triplet interconversion, read out as fluorescence contrast. None of this bears on natural magnetoreception.

Frozen prediction: By 2027-09-05 a journal paper with no authors shared with either Nature 2026 paper reports in-cell ODMR of a flavoprotein radical pair (p=0.5).

DEMOTED viable → modeled. The diagnosis is right and important: the honest Hilbert space of a real cryptochrome radical pair is intractable, so nearly every published 'prediction' rests on a semiclassical approximation known to misestimate exactly the anisotropy the compass claim depends on — which is why the field resists resolution. But 'viable' was claimed for the numerics, and the specific application to radical-pair spin dynamics rests on a 2025 preprint from one group. Tensor-network methods are mature in general; their radical-pair instance is not yet replication-grade. Tracked because…

★ For the kids: To check the bird-compass idea you have to solve a maths problem about tiny spinning magnets. The real problem is too big for any computer, so scientists use shortcuts — and the shortcuts blur the exact detail that matters. Better tools are being built.

Technical: Hilbert dimension 4·∏(2Iₖ+1) reaches ~1e8–1e10 for the full ErCry4a hyperfine set. Hino, Frantzov, Kurashige & Antill, arXiv:2509.22104 (2025): tensor networks (DMRG/TDVP) for radical-pair spin dynamics — preprint. Validated benchmark method: Fay, Lindoy & Manolopoulos, J. Chem. Phys. 154, 084121 (2021). Schulten-Wolynes semiclassics is the usual retreat and the known weak point for directional anisotropy.

Frozen prediction: By 2029-12-31 a published calculation treats ≥20 hyperfine-coupled nuclei of a cryptochrome radical pair without semiclassical approximation, on independently reproducible open code. (Registered as the CALIBRATION COUNTERWEIGHT — expected to resolve TRUE, so the prediction set is not uniformly negative.)

HELD at open with the retirement trigger kept at entry time. Reported magnetic-field dependence of amphibian morphological pigmentation is both light- and eye-dependent — the two hallmarks used to argue for the radical-pair mechanism. The strategic interest is methodological: a developmental/physiological endpoint sidesteps the trained-navigation behavioural noise that wrecked the fly literature. But it is one group, preprint only, unreplicated, and pigmentation carries many light-dependent confounds. Registering the retirement trigger NOW is the guard against the tenure-by-inertia failure mo…

★ For the kids: Frogs and tadpoles can change how dark their skin is, and one team says the change follows the magnetic field — but only with light and working eyes, the same two rules the bird compass follows. One team, one paper. We're writing it down now so we remember to check.

Technical: Zadeh-Haghighi, Bertolesi, McFarlane & Simon, arXiv:2607.20513 (2026). Required before any promotion: intensity- and spectrum-matched illumination, sham-field blinding, pre-registered effect size, independent laboratory.

Frozen prediction: By 2028-12-31 either an independent laboratory replicates the light- AND eye-dependence under blinded sham-field controls, or this component is retired.

HELD at open and firewalled from the animal components. One group reported reproducible alpha-band desynchronisation time-locked to rotations of an Earth-strength field in a shielded chamber, with a polarity dependence argued to rule out induction artefact. Years later there is still no independent replication by a laboratory outside the originating collaboration. A single lab with an excellent protocol is a hypothesis, not a sense — and this is the claim most likely to be over-read from the crew's animal magnetoreception work.

★ For the kids: One team says human brainwaves react to a magnet moving around the room, even though people feel nothing. Nobody else has managed to see it yet — so it stays a maybe.

Technical: Wang, Hilburn, Wu, Deng, Xu, Kirschvink et al., eNeuro 6, ENEURO.0483-18.2019 (2019): alpha-ERD to rotating ~50 uT fields, Faraday-shielded, actively nulled. The reported north-down/inclination dependence is more consistent with a MAGNETITE than a radical-pair mechanism. Chae et al., Sci. Rep. 12, 8997 (2022) used a different paradigm and is not a replication.

Frozen prediction: No independent laboratory outside the originating collaboration publishes a peer-reviewed replication of the alpha-band geomagnetic response by 2029-12-31.

VERIFIED: Zadeh-Haghighi, Siguenza, Smith, Simon, Craddock, Science Advances 12(7):eady8317 (13 Feb 2026). Mg isotope substitution changes tubulin polymerization kinetics in a nuclear-spin-dependent way, enhanced under 3 mT; radical-pair model fitted. One lab, biochemical assay, no cellular replication, 3 mT ~60x geomagnetic. Firewall: zero weight for quantum-consciousness-orch-or.

★ For the kids: Swapping magnesium for a heavier twin with a different nuclear spin changed how fast cell scaffolding fibres grow. Interesting chemistry — not evidence that brains are quantum.

Technical: 25Mg vs spin-zero 24/26Mg is the cleanest RPM discriminator (cannot be mimicked by heating or induced currents); effect size and n need reading from the full text.

Frozen prediction: By 2028-03-05 no second independent laboratory publishes a magnesium-isotope x weak-field interaction on tubulin polymerization (p=0.75).

HELD at modeled. The sharpest single blow to the microtubule-as-quantum-substrate story, and it comes from the superradiance proponents' own work: the cooperative UV effect is predicted across large tryptophan networks generally — amyloid fibrils, virus capsids, other structural assemblies — not uniquely in tubulin. An effect shared with an amyloid plaque and a virus shell cannot be the physical signature of cognition. Capped at modeled rather than viable because the cross-architecture generality is substantially computational rather than measured in each architecture; the burden nonetheless …

★ For the kids: The glowing trick people got excited about in brain scaffolding also happens in virus shells and in the gunk found in diseased brains. If plaque does it too, it can't be the secret of thinking.

Technical: Babcock, Montes-Cabrera, Oberhofer, Chergui, Celardo & Kurian, J. Phys. Chem. B 128, 4035 (2024): superradiant signatures across tryptophan mega-networks in multiple biological architectures — a class result driven by chromophore density and geometry, not tubulin identity. Specificity, not existence, is the load-bearing claim for any cognitive inference, and specificity is positively absent.

Frozen prediction: Cooperative UV emission is reported in at least one additional non-microtubule tryptophan assembly by 2029-12-31, and no paper demonstrates a tubulin-specific enhancement beyond what chromophore geometry alone predicts.

Li et al., BMC Anesthesiol 25:109 (2025; verified) adds bidirectional, drug- and regimen-dependent LORR shifts in one mouse paradigm — the signature of ordinary pharmacology acting on excitability, not a single quantum substrate being dampened. Meyer-Overton / GABA-A / two-pore K+ target pharmacology remains the incumbent.

★ For the kids: Different microtubule drugs made mice fall asleep faster OR slower under the same gas — what you expect from ordinary drug effects.

Technical: Every 'quantum anaesthesia' claim must explain the reversed sign before it explains consciousness.

Frozen prediction: No microtubule-anaesthesia study publishes the receptor-trafficking and partitioning controls needed to exclude the classical baseline by 2028-12-31.

SELF-DEMOTION, viable → modeled: skepticism must be earned with evidence too. The quantitative refutations are real and important — computed tubulin-qubit energetics and Frohlich-condensation regimes both find the specified machinery infeasible on its own terms. But the two flagship calculations come from ONE collaboration, and describing them as 'unrebutted' is inaccurate: the proponents published a book-length response. A single-collaboration critique that has been answered (however qualitatively) is exactly the evidence shape this crew demotes when the conclusion runs the other way. Nothin…

★ For the kids: Other scientists did the actual maths on the machine the theory says is inside brain cells, and the maths says the machine can't run. But it was one team who did that maths, and the theory's authors wrote back. So: a serious objection, not a closed case.

Technical: McKemmish, Reimers, McKenzie, Mark & Hush, Phys. Rev. E 80, 021912 (2009); Reimers, McKemmish, McKenzie, Mark & Hush, PNAS 106, 4219 (2009) — same collaboration. Proponent response: Hameroff & Penrose, Phys. Life Rev. 11, 39 (2014). Independent supporting arguments: Koch & Hepp, Nature 440, 611 (2006); Tegmark, Phys. Rev. E 61, 4194 (2000). Promotion to viable requires a quantitative refutation from an unaffiliated group, or the proponent response being answered in kind.

Frozen prediction: No peer-reviewed paper answers the McKemmish/Reimers energetic objections with a quantitative tubulin-qubit model by 2029-12-31; responses remain qualitative or reformulate the mechanism.

CONFIRMED viable, because the NULL is replication-grade. Real magnetobiology studies microtesla-scale radical-pair chemistry with measured spin physics; the consumer market sells static magnets and 'quantum' devices for pain and wellness, where systematic review and meta-analysis of randomised trials found no benefit. Carried explicitly so the boundary is stated rather than implied — this crew's public magnetoreception material is one careless sentence away from being quotable as endorsement of magnet therapy.

★ For the kids: Shops sell magnetic bracelets that are supposed to make you feel better. Tested properly, lots of times, they didn't work. Real magnet science in birds is a totally different thing.

Technical: Pittler, Brown & Ernst, CMAJ 177, 736 (2007): no effect over placebo. Four discriminators separate the real literature: light dependence, an RF resonance fingerprint at Larmor-matched frequencies, dose-response in field strength and inclination, and an objective rather than subjective endpoint. Consumer devices satisfy none. These same four are adopted as this crew's own promotion criteria for any magnetobiology component.

Frozen prediction: No randomised, adequately-powered, placebo-controlled trial establishes a clinical benefit of static-magnet or consumer 'quantum' wellness devices by 2029-12-31.

HELD at modeled. Error correction protects Clifford gates cheaply, but a universal machine needs non-Clifford T gates injected as magic states and purified, and magic-state factories — not qubit count — dominate the footprint and runtime of every credible fault-tolerant resource estimate. Cultivation protocols cut the cost by roughly an order of magnitude on paper and small logical distillation rounds have run on hardware, but demonstrated logical T error is ~1e-3–1e-2 against the 1e-9–1e-12 that interesting algorithms assume. A ~1e6 gap with no demonstrated scaling path is the hardest-bindin…

★ For the kids: A quantum computer can fix its own mistakes for the easy moves, but the hard move needs a special 'magic' ingredient that has to be purified over and over. Making enough clean magic ingredients is the real hold-up — not building more qubits.

Technical: Bravyi & Kitaev, Phys. Rev. A 71, 022316 (2005); Litinski, Quantum 3, 128 (2019) for space-time accounting in which T-factories dominate a surface-code layout; Gidney, Shutty & Jones, arXiv:2409.17595 (2024) for cultivation. Hardware: small logical distillation rounds on neutral atoms, arXiv:2412.15165 (2024).

Frozen prediction: No peer-reviewed demonstration of a logical T gate or distilled magic state at logical error below 1e-6 on any hardware modality by 2027-12-31.

Tier rests on the 2024-25 journal record (IBM bivariate bicycle; Pecorari et al., Nat Commun 16:1111, 2025 for long-range atom registers). 2026 preprints, all VERIFIED: Tham, Goldman, ... Delfosse (IonQ), arXiv:2606.06455 (4 Jun 2026) — nine codes incl. [[18,4]] qLDPC on one trapped-ion device; 'breakeven' = logical lifetime ~ physical ion lifetime, a weak bar where physical memory is already long; the number to watch is logical error per QEC round vs two-qubit gate error. Nixon, McLauchlan, van Rest (Sydney), 'Vine Codes', arXiv:2606.20263 (Jun 2026) — planar nearest-neighbour qLDPC saves ~2…

★ For the kids: Clever new error-fixing codes promise to need far fewer qubits — but on chips where qubits only talk to neighbours the saving shrinks to about a quarter, and the ion-trap 'breakeven' compares against a raw ion that already remembers for a very long time.

Technical: None of the 2026 items is peer-reviewed. Overhead is decided on the connectivity axis, not the code axis, for superconducting roadmaps.

Frozen prediction: By 2027-12-31 no peer-reviewed trapped-ion qLDPC demonstration reports logical error below 1e-3 per QEC round with >= 4 logical qubits (p=0.6); by 2027-09-05 no journal-published superconducting experiment runs a planar-grid qLDPC code with k >= 4 logical qubits (p=0.8).

HELD at modeled. An error-corrected machine emits syndrome data every ~1 microsecond and must decode faster than it arrives or the queue grows without bound and the machine stalls exponentially. This is a classical-computing problem sitting inside the quantum one, and it is distinct from decoder accuracy: the most accurate published decoders are far too slow to run in the loop. Arguably the most under-priced engineering constraint on the roadmap, and the one with the cleanest picks-and-shovels consequence — every fault-tolerant machine ships with a large FPGA/ASIC co-processor whose count sca…

★ For the kids: The quantum chip shouts out clues about its mistakes a million times a second. A normal computer has to listen and work out the fix faster than the clues arrive — if it falls behind, it never catches up.

Technical: Exponential backlog argument: Terhal, Rev. Mod. Phys. 87, 307 (2015). Parallel/window decoding restores throughput scaling: Skoric et al., Nat. Commun. 14, 7040 (2023). Requirements survey: Battistel et al., Nat. Rev. Phys. 5, 597 (2023). Tension with the ML axis: AlphaQubit (Bausch et al., Nature 635, 834, 2024) is the most accurate reported decoder and runs offline.

Frozen prediction: No peer-reviewed report of a machine-learning decoder running in the real-time control loop (sub-1-microsecond per round, superconducting hardware) at accuracy matching offline matching decoders, by 2027-12-31.

CONFIRMED viable and DEDUPLICATED (two lenses filed this under two slugs). The strongest anti-hype result in quantum computing, and it is a theorem rather than a benchmark: mitigation buys accuracy by paying in sampling overhead, and that overhead is proven to grow exponentially with circuit depth under realistic local noise. Every 'quantum utility before fault tolerance' announcement runs on mitigation, so the ceiling is not an engineering inconvenience that better hardware removes — it is a lower bound. Viable because these are proofs, independently derived and tightened.

★ For the kids: Noisy quantum computers clean up their answers by running the same job over and over and averaging. Mathematicians proved the number of repeats explodes as the job gets longer — so this cleanup trick can never scale to the big problems.

Technical: Takagi, Endo, Minagawa & Gu, npj Quantum Inf. 8, 114 (2022): protocol-independent sampling-cost lower bounds. Quek, Franca, Khatri, Meyer & Eisert, Nat. Phys. 20, 1648 (2024): exponentially tighter depth bounds under local noise. Review: Cai et al., Rev. Mod. Phys. 95, 045005 (2023). Applied instance: Kim et al., Nature 618, 500 (2023). Escapes restrict the noise model or observable class and must be graded as scope reductions, not refutations.

Frozen prediction: No peer-reviewed result exhibits a mitigation protocol with sub-exponential sampling overhead in depth for a generic local-noise model by 2028-06-30.

CONFIRMED viable. The second proven ceiling, under the variational/QAOA/VQE family that most commercial quantum roadmaps sell: gradients vanish exponentially with qubit count, making training infeasible at scale. The vicious twist found later is that the structural conditions which PROVABLY avoid barren plateaus tend to be the same conditions that make the circuit classically simulable. Escape the untrainable regime and you may land in the useless-because-classical regime. Replicated and extended by many independent groups, hence viable.

★ For the kids: To train these quantum programs you follow the slope downhill — but the landscape becomes perfectly flat as you add qubits, so there's no downhill to find. And the ways to keep the slope also let a normal computer do the job.

Technical: McClean, Boixo, Smelyanskiy, Babbush & Neven, Nat. Commun. 9, 4812 (2018): gradient variance concentrates exponentially over unitary 2-designs. Larocca et al., Nat. Rev. Phys. 7, 174 (2025): taxonomy of plateau sources and escapes. Cerezo et al., arXiv:2312.09121 (2023): known plateau-free ansatz families admit efficient classical surrogates.

Frozen prediction: No peer-reviewed demonstration exhibits a variational family simultaneously provably plateau-free at scale AND shown to resist classical surrogate simulation, by 2028-12-31.

CONFIRMED viable: two-qubit fidelities at and above 99.9% have been reported by multiple independent groups on multiple continents, with all-to-all connectivity via ion shuttling, and this is the modality on which most 'logical qubit' firsts were run. The honest counterweight is clock speed: gates are microseconds to milliseconds rather than nanoseconds and shuttling adds more, so a surface-code round on ions is orders of magnitude slower than on transmons. The modality that wins every fidelity benchmark may still lose the fault-tolerance race on wall-clock throughput.

★ For the kids: Ions floating in a vacuum are the most accurate qubits anyone has built — right about 999 times out of 1000. But they're slow, so a machine made of them thinks very carefully and very slowly.

Technical: QCCD: Pino et al., Nature 592, 209 (2021); race-track, Moses et al., Phys. Rev. X 13, 041052 (2023). Fidelities: Clark et al., PRL 127, 130505 (2021); Loschnauer et al., arXiv:2407.07694 (2024) at 99.97% with all-electronic control. Independent groups (Oxford, NIST, Quantinuum, Innsbruck) supply the replication that superconducting below-threshold QEC currently lacks.

Frozen prediction: Trapped ions retain the published record for two-qubit gate fidelity through 2027-12-31, and no trapped-ion system reports a surface-code round time below 10 microseconds in that window.

HELD at modeled. Spin qubits in silicon are micron-scale, operate above millikelvin, and can in principle be made in an existing CMOS fab; single- and two-qubit fidelities crossed the surface-code threshold in 2022 and have been reproduced by independent groups. What has not happened is scale: the largest reported processors remain in the single digits of qubits, and device-to-device variability — set by atomic-scale interface disorder rather than lithography — is the unsolved problem the CMOS argument was supposed to solve.

★ For the kids: These qubits are single electrons parked in silicon, the same stuff normal computer chips are made of — so the world's chip factories could in theory print millions. So far nobody has printed more than a handful that work together.

Technical: Threshold-crossing fidelities: Xue et al., Nature 601, 343 (2022); Noiri et al., Nature 601, 338 (2022); Mądzik et al., Nature 601, 348 (2022) — three independent groups, which is why the FIDELITY claim is solid and the SCALE claim is not. Six-qubit universal control: Philips et al., Nature 609, 919 (2022). Advanced-node fabrication: Zwerver et al., Nat. Electron. 5, 184 (2022).

Frozen prediction: No peer-reviewed silicon spin-qubit processor exceeds 20 simultaneously-controlled qubits with median two-qubit fidelity above 99% by 2028-06-30.

CONFIRMED viable. Transmon coherence improved roughly an order of magnitude in five years with essentially no change to circuit design — the gains came from metallurgy and surface chemistry: tantalum instead of niobium, surface encapsulation, substrate cleaning, reproduced across independent groups. That makes qubit quality a materials problem and puts a physical bill of materials under the whole quantum roadmap: refractory-metal films, isotopically enriched Si-28, high-purity sapphire, and helium-3 for dilution refrigerators. This is the direct handoff to the materials wedge's chokepoint met…

★ For the kids: Quantum chips got much better not by redesigning the circuit but by changing which metal they're made from and how clean the surface is. The recipe now depends on some quite rare ingredients.

Technical: Place et al., Nat. Commun. 12, 1779 (2021) (Ta transmons past 0.3 ms); Wang et al., npj Quantum Inf. 8, 3 (2022) (~0.5 ms); Bal et al., npj Quantum Inf. 10, 43 (2024) (Nb encapsulation); review, Siddiqi, Nat. Rev. Mater. 6, 875 (2021). Loss is dominated by two-level-system defects in amorphous surface oxides. Supply exposure: He-3 is a tritium-decay byproduct with state-controlled stockpiles; Si-28 enrichment capacity is thin.

Frozen prediction: Median reported transmon T1 in peer-reviewed multi-qubit processors (not single-device records) remains below 1 ms through 2027-12-31.

CONFIRMED viable and DEDUPLICATED (filed twice under two slugs). The counter-wave to quantum computing has been repeatedly broken by classical mathematics, on ordinary hardware, after years of public scrutiny: SIKE, a NIST fourth-round candidate, was destroyed on a laptop in an afternoon; Rainbow, a third-round finalist, fell over a weekend. This is the most important base rate for anyone reasoning about migration — the risk to PQC has so far come from classical cryptanalysis, not from quantum computers, and it arrives without warning. Lattice hardness has no reduction to any proven-hard prob…

★ For the kids: Some of the new 'unbreakable by quantum computers' codes turned out to be breakable by an ordinary laptop, before anyone even built the quantum computer. So 'safe' really means 'nobody has found the trick yet'.

Technical: Castryck & Decru, EUROCRYPT 2023 (LNCS 14008): glue-and-split isogeny key recovery on SIDH/SIKE via torsion-point images and Kani's criterion; direct variant, Maino et al. (EUROCRYPT 2023). Beullens, CRYPTO 2022 (LNCS 13508): MinRank-style break of Rainbow in ~53 hours on a laptop. Counter-example in the right direction: Chen's claimed quantum lattice attack (arXiv:2404.16752, 2024) was withdrawn by the author after an independently-found bug — the correct outcome, fast. Consequence: crypto-agility, hybrid key establishment, and family diversity are engineering requirements.

Frozen prediction: No NIST-standardized PQC primitive (ML-KEM, ML-DSA, SLH-DSA, FN-DSA, HQC) suffers a published practical key-recovery break at its claimed security level by 2028-06-30; at least one non-standardized scheme with 3+ years of public scrutiny does.

HELD at modeled. Quantum chemistry is the flagship application argument for fault-tolerant machines, and it is the one where the advantage case has been most seriously attacked from INSIDE the field. Resource estimates for industrially interesting molecules fell several orders of magnitude between 2017 and 2021 through better algorithms; then a 2023 analysis by the same community found no evidence of exponential advantage for ground-state chemistry, because the quantum algorithm needs a good starting state and preparing one is itself hard.

★ For the kids: Everyone says quantum computers will design new molecules. The people who do the maths found a catch: to get the right answer fast, the quantum computer needs a good first guess, and finding a good first guess is just as hard.

Technical: Reiher et al., PNAS 114, 7555 (2017) (~1e14 T gates for FeMoco); von Burg et al., Phys. Rev. Research 3, 033055 (2021); Lee et al., PRX Quantum 2, 030305 (2021) (tensor hypercontraction, ~4 orders off); Goings et al., PNAS 119, e2203533119 (2022) (P450). Skeptic anchor: Lee et al., Nat. Commun. 14, 1952 (2023) — QPE speedup is conditional on ground-state overlap, which decays exponentially for strongly-correlated targets, so polynomial (not exponential) advantage is the defensible claim.

Frozen prediction: No peer-reviewed fault-tolerant quantum computation solves a chemistry or materials ground-state instance to chemical accuracy beyond the best classical result on the same instance, by 2028-12-31.

TIER CONFLICT RESOLVED AT MODELED. Two lenses filed this component at two different tiers — one at modeled, one at viable — and the higher tier is refused. The underground search for the spontaneous X-ray emission that gravity-related collapse must produce is a genuinely decisive and under-cited constraint, and it excludes the PARAMETER-FREE Diosi-Penrose model. But the exclusion rests on one collaboration's underground dataset and its own follow-ups, with no independent replication at another facility; the same single-collaboration standard that pins the human EEG result at open forbids call…

★ For the kids: One wild idea says gravity itself makes quantum weirdness stop. Scientists went deep under a mountain to listen for the faint glow it would have to make and heard nothing — so the simplest version is out. One team, one mountain, so far.

Technical: Donadi, Piscicchia, Curceanu, Diosi, Laubenstein & Bassi, Nat. Phys. 17, 74 (2021): Ge-detector spontaneous-radiation search at LNGS excludes DP with R0 at nuclear scale; surviving variants require a fitted R0 ≳ 1e-10 m (conceded in Derakhshani et al., Phys. Life Rev. 2022). Related CSL narrowing: Vinante et al., PRL 125, 100404 (2020). Positive-evidence counterpart — gravitationally-induced entanglement (Bose et al. / Marletto & Vedral, PRL 119, 2017) — not yet performed. Promotion to viable requires an independent underground facility reproducing the bound.

Frozen prediction: No experiment reports a gravitationally-induced entanglement witness between two masses by 2029-12-31, and the parameter-free Diosi-Penrose exclusion is not overturned in that window.

CONFIRMED viable. Quantum magnetometers have crossed out of physics labs into clinical neuroscience: wearable optically-pumped-magnetometer arrays record magnetoencephalography while the subject moves, replacing helium-cooled SQUID helmets, with multiple independent groups and commercial systems. This is the clearest case of a quantum technology delivering value now, and it belongs in this crew for a second reason — these are exactly the instruments sensitive enough to adjudicate the magnetoreception claims elsewhere in this study. Quantum FOR biology, kept rigorously distinct from quantum IN…

★ For the kids: Special quantum sensors can hear the tiny magnetic whispers your brain makes, from a helmet you can walk around in. The same sensors are sensitive enough to check whether animals really do have a magnetic sense.

Technical: Boto et al., Nature 555, 657 (2018); field review, Brookes et al., Trends Neurosci. 45, 621 (2022). NV-diamond magnetometry resolved single-neuron action potentials (Barry et al., PNAS 113, 14133, 2016). Biomedical scope: Aslam et al., Nat. Rev. Phys. 5, 157 (2023). ~10 fT/√Hz (OPM) and sub-nT at micron resolution (NV) is the instrument class needed to test in-vivo radical-pair effects directly rather than inferring them from behaviour.

Frozen prediction: At least one OPM-MEG system is used in a published multi-site clinical study with n≥50 patients by 2028-06-30.

HELD at modeled. Almost every headline quantum-advantage claim is a sampling task that classical algorithms later caught. One class did not get caught: tasks where a quantum computer holds quantum data in memory and processes many copies coherently, where the separation is PROVEN rather than conjectured and was demonstrated on only tens of noisy qubits. It stays at modeled — not lower — because the proof is real, and not higher because the tasks are contrived learning problems with no demonstrated commercial instance, and 'quantum memory' assumes the data is already quantum. It must be graded…

★ For the kids: Most 'quantum beat the supercomputer' stories got beaten back by cleverer normal computers. One didn't: when the quantum computer is allowed to remember quantum things directly, it really can learn some facts with far fewer experiments — and that one is proved.

Technical: Huang et al., Science 376, 1182 (2022): exponential sample-complexity separations for learning physical-system properties with quantum memory, on 40 superconducting qubits without error correction — the separation is in NUMBER OF EXPERIMENTS, not gate count, so noise does not erase it. Framework: Aharonov, Cotler & Qi, Nat. Commun. 13, 887 (2022).

Frozen prediction: No classical algorithm overturns the Huang et al. (2022) sample-complexity separation by 2028-12-31, and no commercially-motivated instance of it is published in that window either.

Held at open; the two proposals conflicted on a fact and the skeptic verified it. Tindall et al.'s tensor-network work (preprint arXiv:2503.05693, March 2025) was published in Science on 21 May 2026 (Flatiron CCQ + Boston University), matching King et al. (Science 388, 199, 2025) on square, cubic and diamond lattice quench dynamics, with early runs on a laptop. D-Wave's rebuttal ('Quantum Supremacy Result Stands') is factually specific: the hardest 3D high-disorder, higher-order-observable instances were not reproduced. Verdict: advantage retreats to a shrinking corner, the classical-refutati…

★ For the kids: A company said its quantum machine beat all normal computers at a puzzle. A year later a scientist did most of the puzzle on a laptop, and it was printed in a top journal. The company says the hardest corner is still unsolved — which is true, for now.

Technical: Open instances: strongly coupled 3D lattices, high disorder, fast quench, higher-order correlators. Watch for 3D-capable PEPS-BP or sampling methods closing that corner. No third-party hardware re-run exists.

Frozen prediction: By 2027-12-31 a classical method reproduces the 3D high-disorder instances including higher-order observables to stated precision: P=0.55. If yes, the claim joins the refuted list.

HELD at open. QML is the loudest commercial quantum story with the thinnest evidence base, hollowed out by two independent forces: dequantization, where classical algorithms with comparable data-access assumptions match the claimed exponential speedups (showing the speedup lived in the state-preparation assumption, not the quantum processing); and benchmarking hygiene, where head-to-head comparisons against properly tuned classical baselines find no advantage on classical data. There is no replicated advantage result on real data.

★ For the kids: People say quantum computers will learn from data far better than normal ones. But when someone gives the normal computer the same head start, it usually keeps up — the magic was in the head start, not the quantum part.

Technical: Tang, STOC 2019: dequantized recommendation systems via length-squared sampling access. Schuld & Killoran, PRX Quantum 3, 030101 (2022): advantage is the wrong framing; benchmarks lack matched baselines. Huang et al., Nat. Commun. 12, 2631 (2021): data access collapses many putative separations. Bowles, Ahmed & Schuld, arXiv:2403.07059 (2024): no advantage in systematic benchmarks.

Frozen prediction: No peer-reviewed result demonstrates a QML advantage on CLASSICALLY-generated data against a tuned classical baseline that survives 12 months without a dequantization or benchmarking rebuttal, by 2028-12-31.

HELD at open with a deliberately hostile bar, and adopted as the corpus's third arm alongside quantum-IN-biology and quantum-FOR-biology. There is not one quantum device whose measured performance was improved by a biological design principle. Its main function here is early hype capture: proposals to use microtubules as computational substrates land in this bucket, flagged as proposals, and are never evidence for any cognitive quantum claim.

★ For the kids: Engineers copy nature sometimes — planes copied birds, velcro copied burrs. Some people now say quantum computers should copy proteins. Right now that's only an idea. Not one quantum computer anywhere works better because it copied something alive. We're keeping a scoreboard.

Technical: Taxonomy: Gassab et al., arXiv:2605.00205v3 (2026). Candidate templates: tryptophan mega-network UV superradiance (Babcock et al., 2023); collective photoprotection (Patwa et al., Front. Phys. 2024). Mavromatos, Eur. Phys. J. Plus 140, 1116 (2025) proposes microtubules as a scalable computational substrate — recorded as a proposal with zero experimental support and no coherence-time budget surviving Tegmark-class decoherence estimates.

Frozen prediction: By 2030-12-31 no quantum device holds a published, independently reproduced performance record attributable to a biologically derived design principle.

VERIFIED: Nordmann, Balay, ... Malkemper, Landler, Keays, Science (online 20 Nov 2025; print 391:1155, 2026; doi:10.1126/science.aea6425) — whole-brain activity mapping shows light-independent activation of medial vestibular nuclei and caudal mesopallium; scRNA-seq of canal cristae finds type II hair cells expressing 'the molecular machinery' for induction. VERIFIED: Kavokin et al., J R Soc Interface 23:20260129 (1 Jul 2026). Skeptic ruling: Kavokin's paper infers a separate RF-PERTURBATION detector from modulated-vs-CW behaviour in flycatchers; it does not name the inner ear, measure anythin…

★ For the kids: One team thinks pigeons might feel magnetism with the balance organ in the ear. Nobody has switched it off to check, and the second team people point to was studying something else.

Technical: Activation is correlation; no trigeminal signal detected (stimulus choice may matter). Promotion to modeled needs either canal hair-cell ablation with a magnetic-orientation deficit or an independent activity-map replication; viable needs rescue.

Frozen prediction: By 2027-09-05 no peer-reviewed causal manipulation of semicircular-canal type II hair cells produces a magnetic-orientation deficit in any bird (p=0.7).

VERIFIED: P. J. Hore, PNAS 123(8):e2524093123 (Feb 2026). A magnetite particle concentrates 50 uT to ~5 mT so nearby radical pairs read INTENSITY encoding direction; predicted 10–100x sensitivity gain. Theory-only. The 2026 liver-macrophage paper is NOT evidence for this model (no radical-pair partner, superparamagnetic not single-domain).

★ For the kids: A tiny magnet inside the cell could make the chemistry compass much more sensitive — on paper.

Technical: Hybrid-sensor radical pairs sit in the mT regime, so the low-field 'compass intensity window' should differ — a discriminating prediction.

Frozen prediction: By 2028-09-05 no experimental co-localization of magnetite/ferritin particles with a cryptochrome or flavoprotein radical pair in any magnetoreceptive tissue is published (p=0.8).

Held at open. FAD-superoxide tightly-bound pair rescued on paper via asymmetric recombination (Nat. Commun. 15, 2024, s41467-024-55124-x); chirality-bolstered variant is a preprint (arXiv 2505.01519). Both rest on a contested partner radical and a contested CISS mechanism. Two open components multiplied do not make a modeled one.

★ For the kids: A clever idea says even tightly-stuck electron pairs can feel Earth's field if one keeps 'checking' the other. It is maths so far.

Technical: Requires k_S >> k_T (or reverse) by orders of magnitude plus a chiral spin filter; neither measured in any cryptochrome.

Frozen prediction: By 2028-12-31 no measurement of recombination-rate asymmetry at the required magnitude in any cryptochrome radical pair: P=0.8.

A designed carotenoid–porphyrin–fullerene triad shows an orientation-dependent, inclination-type magnetic response at microtesla fields: the physics of the avian compass demonstrated in a bottle at Earth strength. This is the strongest non-biological support the radical-pair hypothesis has, and it sits at OPEN because the entire measurement lineage traces to one collaboration — the identical bar that demoted the honeybee component. The bar is direction-blind, so it must also bite claims the crew finds attractive.

★ For the kids: Chemists built a molecule from scratch that acts like a compass: shine light on it and how fast it reacts depends on how you tilt it in a field as weak as the Earth's. It proves the trick is possible. It does not prove any bird uses it — and only one group has ever made it work.

Technical: Maeda et al., Nature 453:387 (2008) — first anisotropic MFE, far above geomagnetic. Kerpal, Richert, Storey, Pillai, Liddell, Gust, Mackenzie, Hore, Timmel, Nat Commun 10:3707 (2019) — transient absorption on the CPF triad, pronounced orientation dependence in the geomagnetic regime, inclination-compass character. What it does NOT show: any biological molecule, any in-vivo readout, or survival of rotational disorder in a real cell.

Frozen prediction: By 2029-12-31 no group with zero author overlap with the Oxford (Timmel/Hore) collaboration publishes an anisotropic radical-pair magnetic field effect below 100 µT in any molecular system.

HELD at modeled, and it is a deliberate stress test on the 'animal navigation is quantum' narrative. Ansell's mole-rat uses a POLARITY compass (which a radical-pair inclination compass cannot produce), works in total darkness, and its preference flips for months after a magnetising pulse — the magnetite signature. Bats show the same pulse signature from an independent lineage. Enucleation abolishes mole-rat magnetic orientation, implicating corneal magnetite rather than retinal cryptochrome. The single radical-pair-suggestive mammalian result — the wood mouse RF effect — is eleven years old a…

★ For the kids: Mole-rats live underground and have tiny eyes, but they have a magnet sense that works in pitch dark, tells north from south, and can be scrambled by a quick magnet zap. That's the fingerprint of a speck of iron, not quantum spin. One other rodent DID react to radio waves the way a quantum compass should — but only one lab has ever seen it, and nobody has checked in over ten years.

Technical: Marhold, Wiltschko & Burda, Naturwissenschaften 84 (1997); Thalau, Ritz, Burda, Wegner & Wiltschko, J R Soc Interface 3 (2006) — rodent compass light-independent, different physical principle from birds; Caspar, Moldenhauer, Moritz, Němec, Malkemper & Begall, J R Soc Interface 17:20200513 (2020) — bilateral enucleation randomises nest orientation. Independent lineage: Holland, Kirschvink, Doak & Wikelski, PLoS ONE 3:e1676 (2008) — bat pulse remagnetisation. Counterweight: Malkemper et al., Sci Rep 5:9917 (2015) — wood mouse nest axis shifts under 0.9–5 MHz FM field; single lineage, shared sen…

Frozen prediction: By 2029-12-31 no laboratory sharing no senior author with the Duisburg-Essen (Burda/Begall/Malkemper) lineage publishes a replication of radiofrequency-disrupted magnetic orientation in any mammal. If that holds, grade the mammalian magnetic sense CLASSICAL for corpus purposes.

Satellite images of 8,510 cattle and 2,974 deer reported north–south alignment; the only manipulation-controlled test — 34 individually observed cows with and without neodymium magnets — found no directional preference, with sun avoidance explaining the field pattern. Independent labs sit on BOTH sides, so this is capped by DISAGREEMENT rather than by single-lineage weakness, which is a different failure mode and is recorded as such.

★ For the kids: People counted thousands of cows in satellite photos and thought cows face north. Then someone went to the field, stuck magnets on 34 cows and watched: no north, no south — the cows just turned away from the sun to stay cool. Big pictures can fool you; poking the thing is what settles it.

Technical: Begall, Červený, Neef, Vojtěch & Burda, PNAS 105:13451 (2008). Hert, Jelínek, Pekárek & Pavlíček, J Comp Physiol A 197:677 (2011) — independent NULL, rebutted at J Comp Physiol A 197:1127 (2011). Slabý, Tomanová & Vácha (Masaryk), J Comp Physiol A 199:695 (2013) — independent PARTIAL support, conditional on herd density. Weijers, Hemerik & Heitkönig (Wageningen), PLOS ONE 13(4):e0187848 (2018) — the only MANIPULATION study: 97 magnet-collar and 177 control observations plus 2,428 field orientations; null, mean resting orientation ~130° from magnetic north tracking sun position.

Frozen prediction: By 2030-12-31 no peer-reviewed EXPERIMENTAL study using magnet or coil manipulation with individually tracked animals reports statistically significant north–south magnetic alignment in cattle.

Juvenile Pacific salmon and European glass eels tested in coil systems reproducing distant magnetic signatures orient as if they were AT those places, implying a learned or inherited magnetic map. Every paper traces to one research lineage, and the eel result drew a published objection that the animals tested were the wrong life stage. Mechanism is not specified as quantum — magnetite is the working assumption.

★ For the kids: Baby salmon and baby eels put in special magnetic coils act as if they'd been teleported: they swim the direction that would get them home from THERE. One catch — the eels tested were older river eels, not the tiny ocean larvae that make the real journey.

Technical: Putman, Lohmann, Putman, Quinn, Klimley & Noakes, Curr Biol 23:312 (2013); Naisbett-Jones, Putman, Stephenson, Ladak & Young, Curr Biol 27:1236 (2017); Scanlan et al., PNAS 115:10995 (2018) — landlocked population retains the response, arguing inheritance over imprinting. Objection: Durif et al., Curr Biol 27:R998 (2017) — late-stage freshwater glass eels (~2 yr), not pelagic leptocephali, with no evidence one proxies the other; reply at R999 separates map from compass step. Author-overlap concentration across the corpus is near-total (Lohmann lab and trainees).

Frozen prediction: By 2030-12-31 no peer-reviewed magnetic-displacement experiment in any fish showing map-like orientation shifts is published by a group with zero author overlap with Lohmann or Putman.

Three independent nulls and zero independent positives for NATIVE magnetotaxis stand. Burd et al. (Nature 2026) use C. elegans as a host for an engineered radical pair and say nothing about wild-type worms sensing Earth's field.

★ For the kids: Worms given a man-made magnetic protein react to radio waves — that does not mean ordinary worms feel magnets.

Technical: Promotion needs a wild-type behavioural positive from a lab independent of the original claimants.

Frozen prediction: By 2027-09-05 no independent wild-type C. elegans magnetotaxis positive is published (p=0.85).

Abrahams et al., Nature 649:1172 (2026) is the originating group's journal record; Ross, Lodesani & Aiello, bioRxiv 2026.02.18.706690 (verified, posted 19 Feb 2026) characterizes MagLOV2 in E. coli: positive MFE below ~1 mT, sign reversal near 2 mT, plateau above ~70 mT. Independent leg is preprint-tier; the reporter stays open.

★ For the kids: A glow-changing protein has been described in a big journal, and a second team saw the same thing — but their report is not yet fully checked.

Technical: Non-monotonic field dependence with sign reversal is a low-field-effect signature; promote to modeled when the independent leg is journal-published.

Frozen prediction: By 2027-09-05 an independent (non-Oxford, non-Calico) journal paper reports MagLOV/MagLOV2 magnetic-field-dependent fluorescence in living cells (p=0.55).

DEMOTION UPHELD. The claim that a human observer can consciously report a single photon rests on one 2016 study with three observers and overall significance p=0.0545, unreplicated in ten years, and is directly contradicted by an independent 2024 study correlating human psychophysics with primate parasol ganglion-cell output, which concludes that nonlinear ON-pathway processing PRECLUDES perceptual detection of single photons in darkness. AUDIT NOTE THAT CUTS AGAINST THE SKEPTIC: the best-known critique of the 2016 paper was WITHDRAWN by its own authors in 2018 with a concession, and is inadm…

★ For the kids: Your eye's light-catching cells really can react to one particle of light — rock solid. Whether YOU can notice it and say 'I saw that' is different, and a 2024 study says the wiring between eye and brain deliberately throws the single-photon signal away to get better contrast instead. Also, the most famous complaint about the original experiment was taken back by the people who wrote it, so we don't get to use it.

Technical: Tinsley et al., Nat Commun 7:12172 (2016) — 2AFC with heralded SPDC source, N=3, ~30k trials, p=0.0545 overall with significance concentrated in high-confidence trials. Kilpeläinen, Westö, Tiihonen, Laihi, Takeshita, Rieke & Ala-Laurila, Nat Commun 15:4501 (2024) — ON-pathway nonlinearity discards isolated single-photon events while preserving quantal-resolution intensity DIFFERENCES. Not formally incompatible (corneal vs retinal photon accounting, different criterion structures) but pointing opposite ways, with only one mechanistically grounded. Promotion back to viable requires zero-overlap…

Frozen prediction: By 2028-01-01 no independent laboratory (zero author overlap with Tinsley/Vaziri) publishes a replication of human single-photon detection at p<0.01.

Nocturnal bees, deep-sea cephalopods and vertebrate rods all operate where photons arrive one at a time, with performance set by Poisson photon statistics plus thermal isomerisation noise. This is replication-grade quantum-LIMITED biology, and it is the cleanest available demonstration that 'quantum-limited' and 'quantum-coherent' are different claims. Its function in the corpus is as the null model every claimed quantum advantage in photoreception must beat first.

★ For the kids: At night there's so little light that eyes have to count photons one by one. Some bees fly through a dark jungle and still find their nest door, by adding up light from lots of tiny eye-parts and waiting a bit longer between snapshots. Clever biology — but normal physics, not magic quantum physics.

Technical: Shot noise: Warrant et al., Curr Biol 14:1309-1318 (2004); Frederiksen, Wcislo & Warrant, Curr Biol (2008) — single-photon 'bumps' in Megalopta genalis; Warrant, Phil Trans R Soc B 372:20160063 (2017). Solution is spatial and temporal SUMMATION via laterally connecting neurons — a classical trade of resolution for sensitivity, ~30x a diurnal honeybee. Thermal dark noise: Barlow, Birge, Kaplan & Tallent, Nature 366:64-66 (1993); Ala-Laurila et al. (1995); eNeuro 10(11):ENEURO.0390-23 (2023). Activation energies ~23-28 kcal/mol from Schiff-base deprotonation lowering the isomerisation barrier. …

Frozen prediction: Standing falsifier: any claim of quantum-coherent enhancement in photoreception must first beat this shot-noise + thermal-isomerisation null model at matched photon flux and temperature. No such demonstration exists as of 2026-08-01.

The APC non-monotonic curve (Zhou et al., arXiv:2608.23423) is the qualitative fingerprint ENAQT predicts, but engineered ENAQT demonstrations dialled noise at fixed Hamiltonian, whereas temperature changes both bath and vibronic spectrum, and the authors themselves attribute control to spectral-weight distribution rather than dephasing magnitude. 'Proven in analogs, unobserved in biology' becomes 'one candidate observation, preprint'.

★ For the kids: A little jiggling can help energy move — proven in machines, maybe glimpsed in a protein, not confirmed.

Technical: Falsifier: controlled dephasing on APC (deuteration, solvent viscosity) showing the minimum shifts as ENAQT predicts without changing the spectral-weight distribution.

Frozen prediction: By 2028-03-05 no peer-reviewed biological system shows an ENAQT transfer-time minimum reproduced by two groups (p=0.8).

The awareness-adjacent quantum-biology claim closest to actual neurons: that the K+ selectivity filter might exploit coherent transport among backbone carbonyls. Sixteen years, no verification experiment, and the standard structural account (dehydration-energy compensation by precisely spaced carbonyls, knock-on conduction) explains the selectivity without it. The one substantive follow-through found that CLASSICAL coherence plays no role — removing the classical version of the mechanism without supplying the quantum one. Near-retirement is a live question for the next audit.

★ For the kids: A potassium channel is a tiny gate in a nerve cell that lets one kind of atom through and blocks another almost perfectly. In 2010 two physicists asked whether it uses a quantum trick and said how to check. Sixteen years later nobody has checked, and the ordinary chemistry explanation works fine. Asking a good question isn't answering it.

Technical: Vaziri & Plenio, New J Phys 12:085001 (2010) — coherent coupling among backbone C=O modes (~1650 cm^-1), verification proposed via 2D IR of the filter. Salari V. et al., J Integr Neurosci 14(2):195-206 (2015) — MD computes classical coherence lifetimes too short and too weakly correlated to affect conduction. Absence of the classical version is not evidence for the quantum version. RETIREMENT TRIGGER: if by 2027-08-01 no 2D-IR or equivalent test of filter coherence is published, move to the same near-retirement status as quantum-olfaction.

Frozen prediction: By 2027-08-01 no peer-reviewed experimental test (2D-IR or equivalent) of coherence in the KcsA selectivity filter will have been published.

Living tissue really does emit 10–1000 photons per cm² per second and it can now be imaged. But the same group that imaged it has since shown that reported extracranial 'brain biophoton' signals are explained by background light contamination. Carried as a FIREWALL entry: UPE is ordinary ROS chemiluminescence, and is NOT evidence for optical neural signalling, microtubule communication, or anything in the awareness family. The self-refutation by the original proponents is exactly the behaviour this ledger should reward and is why the domain is admitted at all.

★ For the kids: Every living thing glows a tiny bit — far too faint to see, but a very good camera can. It comes from ordinary chemistry inside cells, like very slow rusting. It is NOT the brain sending secret light messages: when scientists checked carefully, that turned out to be stray light leaking into the room.

Technical: Positive: Salari, Seshan et al., J Phys Chem Lett 16(17):4354-4362 (2025) — EMCCD/CCD single-photon imaging (>90% QE), live-vs-dead contrast in mice, injury/temperature-driven increases in plants; mechanism is excited carbonyls and singlet oxygen, i.e. chemiluminescence, not a coherent optical channel. Negative, same group: Salari, Seshan, Rishabh, Oblak & Simon, arXiv:2603.26630 (2026) — large reported head signals explained by background contamination, short-wavelength photons strongly attenuated by scalp and skull. ADMISSIBILITY RULE: no UPE result may be cited in support of microtubule-qu…

Frozen prediction: By 2027-12-31 no independent laboratory publishes a peer-reviewed replication of extracranial biophoton detection correlated with neural activity that survives a pre-registered light-contamination control.

FLAGGED FRONTIER, AWARENESS UNASSERTED. A 2024 paper models the myelin sheath as an optical cavity and calculates that cascade emission from C–H vibrational modes could produce entangled photon pairs, suggested as a neuronal synchronisation resource. It is a THEORY paper: no photon measured from myelin, no entanglement measured, nothing in living tissue. The August 2024 press cycle framed it as explaining consciousness — a claim the paper does not make and the data do not support, because there are no data.

★ For the kids: Some scientists did a maths calculation saying the fatty tube around a nerve could work like a tiny mirror-box making pairs of linked light particles. Nobody has ever seen one. The newspapers wrote 'this explains consciousness'. A calculation is a good guess about where to look — it isn't a thing anyone found.

Technical: Liu Z., Chen Y.-C. & Ao P., Phys Rev E 110:024402 (2024). Unaddressed: thermal occupation of a ~2900 cm^-1 mode at 310 K is negligible so a pumping mechanism is required and none is specified; myelin is a strong optical scatterer; no in-vivo detection scheme with the required dark-count floor exists. Promotion to modeled requires an independent theory group reproducing the emission rate with realistic optical loss; promotion to viable requires a direct coincidence measurement of photon pairs from myelin ex vivo with a classical-light null control.

Frozen prediction: By 2028-01-01 no peer-reviewed paper reports a direct coincidence-counting measurement of entangled photon pairs emitted from myelin or from C–H vibrational modes in any biological tissue.

A firewall component. Quantum cognition borrows quantum probability mathematics — superposition, non-commuting measurement, interference — to fit judgement data such as order effects and the conjunction fallacy. It makes NO claim that the brain runs on quantum physics, and its founders say so explicitly; so do leading Orch-OR advocates, who note quantum-like statistics emerge in complex CLASSICAL systems under coarse-graining. MATURITY NOTE, load-bearing: 'modeled' scores the descriptive fit of the formalism to behavioural data ONLY. Any physical reading — that Hilbert-space fits evidence qua…

★ For the kids: Some scientists borrow quantum maths to describe how people make choices — because the maths fits, the way one equation describes both a swing and a guitar string. That does NOT mean your brain is a quantum computer, any more than a swing is a guitar. Even the scientists who most want brains to be quantum agree about this.

Technical: Pothos & Busemeyer, Behav Brain Sci 36(3):255-274 (2013); Busemeyer & Wang, Curr Dir Psychol Sci 24(3) (2015). The parameter-free QQ equality (Wang, Solloway, Shiffrin & Busemeyer, PNAS 111(26), 2014) is a genuine a priori prediction confirmed across many national survey pairs — that is why the descriptive tier is modeled and not open. Concession against interest: Wiest M.C. & Gupta D.S., Front Hum Neurosci (2026), doi:10.3389/fnhum.2026.1838245, state mental entanglement does not necessarily imply a genuinely quantum physical implementation. PROMOTION IS CAPPED: this component can never exce…

Frozen prediction: By 2028-01-01 no peer-reviewed paper derives a quantitative quantum-cognition model parameter from an independently measured neurophysiological quantity.

Held at modeled; promotion to viable refused. The gate result is verified and is a genuine number-plus-code-type-plus-device-plus-group milestone: a Swap-Wait-Swap controlled-phase gate between dual-rail cavity qubits via a tunable transmon coupler, ~99.9% fidelity at ~500 ns, with photon loss converted to flagged erasures. The tier holds rather than rises because it is one group, one device, and no independent replication of the erasure-gate combination. Erasure conversion as a CLASS has multi-platform support, which is what keeps this at modeled rather than open; the specific dual-rail cavi…

★ For the kids: A quantum bit built from two tiny microwave boxes can raise its hand when it loses information, and now two of them can work together on a step of a calculation. One team did it once. The 'ten times better' number people quote is from a computer simulation, not a measurement.

Technical: 'An entangling gate for dual-rail erasure qubits', Nature 656:47-53 (2026), doi:10.1038/s41586-026-10822-y (PubMed 42557378; Quantum Circuits / D-Wave, Schoelkopf lineage) — verified. Class support from other lineages: Teoh et al., PRX Quantum 4:040303 (2023); Levine et al., PRX Quantum 5:010337 (2024), dual-rail transmon; Ma et al., Nature 622:279 (2023), neutral-atom erasure conversion; Wu, Kolkowitz, Puri & Thompson, Nat Commun 13:4657 (2022), threshold benefit. Metrological caution: fidelity conditioned on no erasure flag is not the same quantity as unconditioned fidelity, and the two are…

Frozen prediction: By 2028-08-25, P=0.35 that a group with no Yale/Quantum Circuits/D-Wave co-authors publishes a dual-rail cavity two-qubit gate with erasure detection at ≥99.5% fidelity in a peer-reviewed venue.

Encoding logical information in the grid states of a single harmonic oscillator. Break-even error correction is replicated and now extended to qutrits and ququarts, and a universal single-logical-qubit GKP gate set exists — but no error-corrected two-qubit GKP logical algorithm has been shown, so the architecture claim stays modeled. Break-even is a memory statement, not a computation statement.

★ For the kids: Instead of lots of little switches to store one safe answer, this stores the answer as a pattern of ripples inside a single box of light — drawn on a grid, so you can tell when it has slipped.

Technical: Sivak et al., Nature 616:50-55 (2023) — real-time GKP QEC beyond break-even. Nature (2025), doi:10.1038/s41586-025-08899-y — qudit break-even, logical qutrit and ququart gains 1.82±0.03 and 1.87±0.03. Nat Phys (2025), doi:10.1038/s41567-025-03002-8 — universal gate set for GKP logical qubits. Grid-state SPAM <1e-3 (arXiv:2607.06718, 2026) is a PREPRINT pointer carrying no tier weight. The unmet requirement is the one that keeps every bosonic approach at modeled: concatenation into a multi-logical-qubit machine running an error-corrected algorithm.

Frozen prediction: By 2027-12-31 no peer-reviewed journal paper reports a two-qubit LOGICAL gate between GKP-encoded qubits operating above break-even.

Every logical-qubit roadmap assumes millions of physical qubits, and every physical qubit needs control wires and electronics. Dilution-refrigerator cooling power is a hard physical budget and current cryo-CMOS control dissipates roughly three orders of magnitude too much per qubit. A falsifiable engineering ceiling no qubit-count roadmap prices in. The ~1000x deficit is a review-level ESTIMATE, not a measurement, and is graded forward as such.

★ For the kids: A quantum computer has to be colder than outer space. Every wire into the cold box carries a little heat. A million qubits means a million wires — and that much heat would cook the computer. Nobody has solved this.

Technical: Xue et al., Nature 593:205-210 (2021) — CMOS-based cryogenic control of silicon quantum circuits. Nature (2025), doi:10.1038/s41586-025-09157-x — spin-qubit control with a milli-kelvin CMOS chip below ~70 mK. TU Delft/Bluefors cryo-CMOS DAC + demultiplexer + crossbar, IEEE Trans Quantum Eng (2025) — hundreds of devices from a few room-temperature lines. Binding constraint is thermal: commercial pulse-tube cryostats deliver ~2-4 W at 4 K, implying an upper bound of order a few microwatts per qubit for a million-qubit control plane. This is the fault-tolerance analogue of materials-wedge chokep…

Frozen prediction: By 2028-12-31 no peer-reviewed paper reports a cryogenic control system driving ≥100 qubits at ≥99.9% single-qubit gate fidelity with total control-electronics dissipation inside the published cooling budget of a standard commercial cryostat.

Quantum processors are simulating confinement, string breaking and glueball dynamics in lattice gauge theories — genuine peer-reviewed physics, and the strongest SCIENTIFIC case for near-term quantum simulation. What has not been shown is that any of it was beyond a good classical tensor-network calculation, which is exactly where the classical-refutation pattern keeps striking. The value proposition is 'sign-problem-free real-time dynamics', not 'classically impossible'.

★ For the kids: Quantum computers are being used to watch how the glue holding atoms' cores together stretches and snaps — like watching an elastic band break in slow motion. The physics is real. Whether an ordinary computer could have worked out the same answer, we don't know.

Technical: Nat Phys (2025), doi:10.1038/s41567-024-02723-6 — confinement in a Z2 LGT. Nat Phys (2025), doi:10.1038/s41567-025-02797-w — 2D lattice QED on trapped-ion qudits via VQE. Nature (2026), doi:10.1038/s41586-026-10445-3 — digital quantum magnetism. Preprint pointers only, no tier weight: arXiv:2604.07435, arXiv:2601.16166. Tier is modeled because the classical baseline is a moving target — MPS and neural-network quantum states have repeatedly caught up within 12-24 months. Promotion condition: a result that survives ≥18 months of documented classical attack.

Frozen prediction: By 2028-12-31 at least one headline lattice-gauge quantum-simulation result from 2025-2026 is reproduced classically (tensor-network or neural-network quantum states) in a peer-reviewed paper.

Proposed promotion to viable REJECTED. Two independent Nature 2025 papers (USTC, Zhou et al., s41586-025-09790-6; Princeton/Berkeley, Rovny et al., s41586-025-09760-y) are convergent evidence that entanglement between NV pairs is a usable metrological resource at ambient conditions — but by the proposer's own text they use DIFFERENT protocols measuring DIFFERENT quantities (3.4x sensitivity and 1.6x resolution vs a change in readout-noise scaling from quadratic to linear). Two groups reaching a shared qualitative conclusion by different routes is strong support for 'modeled'; it is not replic…

★ For the kids: Two teams on two continents both found that linking two diamond sensors helps them hear faint magnetic whispers. They measured different things in different ways, though — so it is strong evidence, but not yet the 'same experiment done twice' that we need to call it settled.

Technical: Convergent-vs-replicated is the distinction: USTC's gain is a signal-interference/noise-suppression gain; Princeton's is a readout-noise-scaling gain that only matters in the readout-limited regime. A replication would be a third group recovering either stated factor. Until then the physical effect is robustly plausible and once-per-protocol demonstrated.

Frozen prediction: By 2028-08-21 either (a) a third group reproduces the USTC 3.4x-class sensitivity gain or the Princeton readout-scaling change on the same protocol, or (b) an entangled multi-NV sensor resolves a target the same group cannot resolve with single NVs. If either, promote to viable. If neither, hold modeled. Confidence neither lands: 0.45.

MERGED (submitted three times across crews, at two different tiers; unified at the lower one). Three claims must be kept apart. The 6 dB bound on the error-probability EXPONENT is a theorem and is solid. The laboratory effect is measured — ~1 dB at 1 m, and ~0.79 dB against a 2.8 dB configuration ceiling in the most careful microwave experiment — by two independent groups. The OPERATIONAL RADAR is refuted and is retired outright: the advantage requires single-photon-level brightness, a preserved quantum-limited idler memory, and prior knowledge of target range, which confines the protocol to …

★ For the kids: People said a quantum radar could see invisible planes. Physicists worked out the very best it could ever do: about four times better at one specific job, only when the signal is basically single specks of light, and only if you already know how far away the thing is. A normal radar just shouts louder. Sometimes the exciting story is over before it starts, because someone already did the maths.

Technical: Lloyd, Science 321:1463 (2008); Tan et al., Phys Rev Lett 101:253601 (2008) — 6 dB Chernoff-exponent bound at equal transmitted energy. Barzanjeh, Pirandola, Vitali & Fink, Sci Adv 6:eabb0451 (2020) — ~1 dB, free space, 1 m. Assouly, Dassonneville, Peronnin, Bienfait & Huard, Nat Phys 19:1418 (2023) — ~0.79 dB against a 2.8 dB ceiling. Shapiro, IEEE Aerosp Electron Syst Mag 35(4):8-20 (2020) — an originator delineating the narrow advantage regime. Karsa, Fletcher, Spedalieri & Pirandola, Rep Prog Phys 87:094001 (2024) — proponents conceding 'aspirational conjecture as well as criticism'. Rang…

Frozen prediction: By 2029-12-31 no peer-reviewed paper demonstrates a microwave quantum-illumination system detecting a target beyond 100 m with a measured advantage over a classical system of matched transmitted energy and bandwidth, and no NATO-country defence programme announces an operational quantum radar.

If two masses in spatial superposition become entangled through their mutual gravity alone, gravity cannot be a classical field. The proposal is well-posed and the theory literature is very active — and NO experiment has produced gravitationally induced entanglement. Theory-only, and it must not drift. Note the honest asymmetry: a null result does not prove gravity is classical, only that sensitivity was insufficient, so this domain can be promoted by a positive result but can never be retired by a negative one — which is itself a reason to keep it at open indefinitely.

★ For the kids: Everything else we know — light, magnetism, atoms — is quantum. Gravity might not be. Scientists want to hang two tiny weights so each is in two places at once, and see if gravity alone can tie them together. Nobody has managed it.

Technical: Bose–Mazumdar–Vedral / Marletto–Vedral witness. Requirements: superposition separations 10-100 µm, masses ~1e-14 kg, second-scale coherence, Casimir and EM screening many orders of magnitude beyond current levitated-optomechanics performance. Theory pointers only, all preprint/PRD-level and none experimental: arXiv:2502.12474 (2025); arXiv:2503.03585 (2025); Phys Rev D (2026) parametric-resonance and operational-QFT GIE papers. Shares hardware with gravitational-collapse-model-tests but asks the opposite question and is graded separately.

Frozen prediction: By 2030-12-31 no peer-reviewed experiment reports a statistically significant gravitationally induced entanglement witness between two masses.

'Quantum battery' has entered energy-technology press as an imminent charging technology. The physics of collective, faster-than-linear-in-N charging is genuine and has proof-of-principle demonstrations in organic microcavities. What does not exist is a device: stored energies are many orders of magnitude below any practical threshold, the systems need cryogenic or tightly engineered optical environments, and no discharge pathway into a useful load with the advantage preserved has been demonstrated. The load-bearing public claim has no experimental support at all — only a scaling argument.

★ For the kids: Line up lots of tiny quantum things and they can soak up energy faster as a team than one by one. Real and cool. But the amount of energy is unbelievably tiny — nowhere near charging a phone — and nobody has built anything you could plug in. 'Real effect' and 'real product' are two different sentences.

Technical: Ferraro D., Cavaliere F., Genoni M.G., Benenti G. & Sassetti M., Nat Rev Phys 8:115 (2026), doi:10.1038/s42254-025-00906-5 — an overwhelmingly theoretical field with a small number of proof-of-principle demonstrations. Superextensive charging (power ~N^a, a>1, Dicke-type coupling) demonstrated in organic microcavities (Quach et al., Sci Adv 8:eabk3160, 2022) at picojoule-to-nanojoule stored energies. Open problems: ergotropy extraction into a work reservoir, storage decoherence, and whether the advantage survives once charger-preparation cost is counted.

Frozen prediction: By 2028-01-01 no peer-reviewed demonstration reports a quantum battery storing ≥1 mJ with a measured superextensive advantage over a classical charging baseline including charger-preparation cost.

DEMOTED modeled→open. HNDL is the load-bearing justification for every PQC migration deadline, and the component's entire analytical support is ONE unrefereed preprint presenting an economic model — no measurement, and zero public evidence that any real HNDL corpus exists or has been decrypted. A single preprint modelling exercise cannot carry modeled; that is the same bar that keeps the hybrid compass, the quantum-Zeno rescue and the myelin biphoton calculation at open, and it is direction-blind. The threat model is also structurally unfalsifiable in the short run, which is a reason for a na…

★ For the kids: Some people worry spies are recording secret messages TODAY so a future quantum computer can unlock them LATER. Saving messages is cheap, so that part is easy for them. The clever defence is changing the lock often, which makes the future unlocking job much more expensive. But nobody has ever caught anyone doing this, and the whole argument so far is one unfinished paper's maths.

Technical: Blanco-Romero, Almenares Mendoza, García Rubio, Campo & Díaz Sánchez, arXiv:2603.01091 (2026) — PREPRINT, open-source testbed over TLS 1.2/1.3, QUIC, SSH. Its useful asymmetry argument: storage inflation (via Encrypted Client Hello) penalises defender and attacker alike, whereas quantum-workload inflation via rekeying penalises only the attacker — so rekeying policy, not key size alone, is the strongest lever. DEMOTION GROUNDS: single preprint, economic model not measurement, no observed instance of the attack. The deployment context (Cloudflare telemetry) is verified but belongs to pqc-real-…

Held at open. One counted in-house case (Quantum Echoes: 1 refutation-relevant paper, 0 independent, ten months out) and one counted independent case (D-Wave annealing: Flatiron/BU Science 2026, no D-Wave authors, ~14 months out). Two anecdotes do not make a ratio; the component's own standard forbids promotion. Deliverable: the independent-refutation ratio across four claim families (see studies).

★ For the kids: If the only person who has raced you is your teammate, you have not proved you are fastest. We are now counting how often that happens.

Technical: Metric: independent attempts (no shared affiliation with claimant) / total attempts within 24 months, per claim family (RCS, GBS, annealing, OTOC).

Frozen prediction: By 2027-08-21 the ratio across the four families for post-2023 claims is below 0.5: P=0.6 (lowered from 0.65 — the D-Wave case is an independent erosion and cuts against the thesis). Falsified at or above 0.5; retire the thesis as overstated.

Case A (verified): Wiest, Neurosci Conscious 2025 niaf011 asserts 'direct physical evidence of a macroscopic quantum entangled state in the living human brain' sourced to a contested, unreplicated J Phys Commun paper. Case B (verified): same-lab mouse follow-up (Huang 2026) circulates as confirmation while the actual independent study (Li 2025) reversed the sign. Case C (verified in outline): Microsoft's Nature Reply (2026, doi:10.1038/s41586-026-10568-7) concedes coding errors in the topological-gap-protocol analysis and argues the parity-readout case does not rest on that data — an evidenti…

★ For the kids: Being printed in a science journal does not make a claim true — journals also print the over-excited versions.

Technical: 'The Quantum Hype in the Scientific Sphere', Digital Society (2026), doi:10.1007/s44206-026-00269-z — STS expectation framework applied by qualitative content analysis to highly cited quantum-technology papers 2007-2023; identifies USE CASE and FUTURE PROJECTION framing modes. Author list not verified by the submitting crew; title, journal, year and DOI verified. DEMOTION GROUNDS: single qualitative study, purposive sample, no reported inter-coder reliability. Adjacent support exists for the general proposition but not for this specific finding: Hoefler, Häner & Troyer, Commun ACM 66(5):82-87…

Frozen prediction: By 2027-08-01 no replication of the Digital Society expectation-framing analysis with an independent sampling frame and reported inter-coder reliability is published.

Count update, all primary and verified: inner-ear induction (Nordmann, Science, Nov 2025); Hore's hybrid magnetite–radical-pair sensor (PNAS 123(8):e2524093123, 24 Feb 2026); liver superparamagnetic macrophages (Lisowski, Science, May 2026); a separate induction-based RF-perturbation sense (Kavokin, JRSI, Jul 2026). Lam & Malkemper, Physiol Rev 106:2383 (2026) still list three main hypotheses. The count is external; the 'degenerating programme' verdict remains ours.

★ For the kids: In under a year scientists added four new ideas for how birds feel magnetism. More ideas is not more answers.

Technical: Candidate avian mechanisms with a named tissue now >= 6. Any two being shown to be the same system would REDUCE the count.

Frozen prediction: By 2027-09-05 the number of peer-reviewed proposed avian magnetoreception mechanisms/variants with a named tissue grows by at least one more (p=0.6).

Held at modeled. Zhou et al. (Nature 2025) on up to 448 neutral atoms: correlated decoding across rounds keeps exponential suppression in d while time overhead drops from d rounds to O(1); companion architecture paper s41586-025-09848-5. Single lineage (Harvard/MIT/QuEra/Yale), primitives and simulation, the 10-100x runtime advantage is a projection. Space overhead (magic states, qLDPC) untouched.

★ For the kids: Instead of checking your homework after every sum, do the page and check once — and still catch every mistake. Only a few pages have been done this way.

Technical: Transversal gates correlate errors across blocks; joint decoding digests the whole syndrome history. Saving is in TIME, not qubit count — does not move the CRQC qubit clock.

Frozen prediction: By 2028-08-21 no independent group publishes an end-to-end logical algorithm with correlated decoding at d>=7 showing >=10x measured wall-clock reduction vs lattice surgery on the same hardware: P=0.7. If one does, promote to viable.

Held at modeled. Nature 2025 (10.1038/s41586-025-09596-6): 300,000 atoms/s reloaded via optical-lattice conveyors, >30,000 initialised qubits/s, >3,000-atom array held >2 h with stored qubits coherent. Single lab, single paper, idle storage only — not sustained QEC. The load-bearing claim is isolation of stored qubits during reload, untested under active syndrome extraction.

★ For the kids: A machine of thousands of laser-held atoms used to leak atoms like a bucket. A conveyor belt now tops it up faster than it leaks — two hours instead of seconds.

Technical: Array size and run duration are currently traded (Caltech 6,100 atoms <13 s vs this 3,000 atoms >2 h); nobody has both.

Frozen prediction: By 2027-08-21 a peer-reviewed paper reports >1 h continuous operation WHILE running repeated stabiliser measurement on a logical qubit: P=0.45. If every run remains idle-storage-only, demote to open.

Held at open. Waterloo, PRX 2026 (10.1103/c994-bzx7): trityl-OX063 single molecule on ~100 nm nanowire, 400 us coherence (~60x spin echo, part of it from a new decoupling sequence), ~10 nuclear spins detected. One lab, one paper; single-protein imaging explicitly not achieved. 'Quantum FOR biology' — says nothing about biology being quantum.

★ For the kids: Old quantum sensors were stuck inside a diamond, like feeling a coin through a thick glove. This one has no glove — but can only feel about ten atoms at once.

Technical: Trades NV optical readout and room-temperature robustness for standoff distance and mechanical readout; the 60x needs an independent referee since it is against standard spin echo.

Frozen prediction: By 2029-08-21 no peer-reviewed single-protein structure or conformational trajectory from any molecular or NV spin sensor: P=0.75. If one appears, promote to modeled.

Held at modeled. USTC, Science 2026 (10.1126/science.aec6243): single Rb atoms, frequency conversion to telecom, positive asymptotic rates to 100 km; at 11 km, 624 h for 1.2M heralded pairs at 0.112 bits/event finite-size. DI-QKD as a physical protocol has prior independent metre-scale demonstrations (Oxford ions, Munich atoms, USTC photons, 2022), which is why modeled rather than open; the 100 km result is single-lineage. DI closes device trust, not the seed/no-signalling/post-processing assumptions.

★ For the kids: Quantum secret codes usually need you to trust the box that made them. This one works even if a stranger built the box — but it took 26 days to make one tiny key.

Technical: Gap to any operational key rate is many orders of magnitude; this is physics, not a product.

Frozen prediction: By 2030-08-21 no DI-QKD system with finite-size secure key rate above 1 bit/s over >=50 km: P=0.7. If one appears, promote to viable and re-open the certified-randomness ledger.

Held at open, and the two-sidedness is why. Pulling for: a radical-pair model predicted a U-shaped superoxide response to field strength in regenerating planaria and the measurement matched — the corpus's best instance of prediction-then-test in this lens, though prediction and test share a collaboration. Pulling against: a structure-resolved analysis of mitochondrial complex I, the most-cited candidate site for spin-dependent ROS chemistry, finds the spin-sensitive region is a narrow shell whose effect magnitude collapses as spin dephasing gets faster, so observable effects would require unu…

★ For the kids: Cells make a reactive chemical called superoxide, and magnets seem to change how much. One experiment matched its own prediction. Another study of where this would have to happen inside a cell says the spot is tiny and the timing would have to be unusually lucky.

Technical: Rishabh R et al., PNAS Nexus 5(5):pgag152 (2026), doi:10.1093/pnasnexus/pgag152 (verified). Sung J, Antill LM, Cheong J, bioRxiv 10.64898/2026.07.15.738821 — preprint, carried as constraint not evidence. Yoon S et al., RSC Adv 16:25395-25401 (2026), doi:10.1039/d5ra10059c — flagged for inadequate thermal/convective and electrode artifact controls. Adjacent and already carried at the same single-lab, mechanism-by-inference tier: Comput Struct Biotechnol J (2025), doi:10.1016/j.csbj.2025.11.055, bell-shaped low-field stimulation of cardiac mitochondrial bioenergetics. Instrumentation that could…

Frozen prediction: Carried by planarian-superoxide-rp-verification.

Held at open. Beutner et al., Comput. Struct. Biotechnol. J. 30:144-157 (2025): up to 40% State 3 respiration increase near 4.6e-4 T, inverted-bell mROS, Complexes II/III/V respond, I does not. Isolated rat-heart mitochondria, single lab, radical-pair attribution is the authors' inference from response shape; the cheap falsifiers (RF resonance, magnetic-isotope substitution) were not run.

★ For the kids: Tiny power plants in heart cells worked harder inside one exact magnet strength and worse in a stronger one. That fingerprint might be electron spins — one lab has seen it once.

Technical: Non-monotonic windows are also produced by classical mechanisms (competing enzyme kinetics, magnetophoresis, oxygen-tension coupling).

Frozen prediction: By 2028-08-21 an independent lab replicates a non-monotonic window in isolated mitochondrial respiration: P=0.35. No replication: RETIRE. Any published RF-resonance null on this preparation: retire immediately.

Held at open, carried as NEGATIVE transfer evidence. Imperatore, Asbury & Giebink, J. Chem. Phys. 154:191103 (2021) reproduced the Rabi splitting and found no rate change in cyanate hydrolysis; Wiesehan & Xiong, J. Chem. Phys. 155:241103 (2021) similar; Annu. Rev. Phys. Chem. 2025 review: collective coupling to 10^6-10^10 molecules gives 1/sqrt(N) per-molecule suppression, disorder suppresses further. Same objection applies to tryptophan superradiance and microtubule photophysics claims.

★ For the kids: Put molecules in a tiny mirror box and some say chemistry changes with the lights off. Others built the same box, saw the same mirror signature, and the chemistry did not change.

Technical: Collectivity does not localise into single-molecule reactivity; a specific quantitative reason to expect nulls.

Frozen prediction: By 2028-08-21 no VSC rate modification reproduced by three or more independent labs on the same reaction with pre-registered detuning controls: P=0.85. Triple replication promotes to modeled and raises the prior on ENAQT-class biology claims.

Held at open, merged from three overlapping proposals, and materially corrected. Two proposers asserted specific milestone figures — a '49-qubit system at 20x' and a '181-qubit system at 2,000x' — and one asserted an SEC accession number. None appears anywhere in the local EDGAR corpus; all are struck as unsupported, and the predictions built on them are rewritten. What the primary filings DO say is verified verbatim below. The separation the proposers got right is kept: the Nature gate paper is credited to erasure-conversion-qubits; this component tracks only the corporate claims wrapped aro…

★ For the kids: A company bought the team that makes the hand-raising qubit and promised machines by 2030 and 2032. The science paper behind it is real. The promises are promises, and we wrote down the dates so we can check.

Technical: Verified in (engine record): EX-99.1 filed 2026-01-07 (D-Wave Quantum Inc., CIK 0001907982, accession 000190798226000010) — acquisition of Quantum Circuits Inc. for $550M, $300M in D-Wave common stock plus $250M cash; 'Quantum Circuits' dual-rail technology with built-in error detection results in higher quality qubits and dramatically lowers the physical resources required for building logical qubits'; 'The first deliverable in the accelerated roadmap will be an initial dual-rail system, planned to be generally available in 2026.'…

Frozen prediction: By 2026-12-31, P=0.45 that D-Wave makes a gate-model dual-rail system generally available to external users (not a demo). By 2026-12-31, P=0.12 that a peer-reviewed or arXiv paper on that platform reports a measured logical error rate below the best physical error rate on the same device. Both graded against the EDGAR record and public cloud access.

Kavokin et al. applied the discriminator and found shape wins. Still a method with one lineage; held at open until an Oldenburg-type lab runs it with equal-mean-power AND equal-peak-amplitude controls plus a modulation-frequency sweep.

★ For the kids: A test that asks whether radio noise matters because of how strong it is or because of its rhythm. So far rhythm — one team has checked.

Technical: Spec: CW vs modulated at equal mean power and equal peak amplitude; 50 Hz–5 kHz modulation sweep; blind funnel scoring; pre-registered n.

Frozen prediction: By 2027-09-05 the discriminator has been run by at most one research lineage in journal literature (p=0.65).

Burd et al., Nature (2026) extend the Stanford RFP-flavin system (JACS 2025, doi:10.1021/jacs.5c03997) into transgenic C. elegans. Same lineage; single-lab rule holds. Double-count guard: engineered transgenic result, not evidence for native celegans-magnetotaxis.

★ For the kids: The same team that found a protein whose glow reacts to magnets put it into a worm and steered it with radio waves — still one team.

Technical: In vivo ODMR contrast is smaller than in vitro; coherence-time bound (>~4 ns) from RF linewidth.

Frozen prediction: By 2027-09-05 no lab independent of the Stanford lineage publishes a >10% static magnetic-field effect on RFP-flavin fluorescence in living cells (p=0.6).

Held at open, as proposed. This is the corpus's cleanest instance of write-the-prediction-then-measure-it in quantum biology: the model predicted superoxide elevated at BOTH ends — hypomagnetic and above ~500 µT — and the measurement matched. Two things keep it at open, one of which the proposers had right and one of which they missed. Right: prediction and test come from the same Calgary-theory/Western-Michigan-experiment collaboration, so this is an internally closed loop, not independent verification; the authors themselves report the blastema-size claim did not hold and the radical identi…

★ For the kids: Flatworms regrow their bodies. Scientists predicted magnets would change a chemical inside them in an unusual U-shape, measured it, and were right. But the thing they thought it explained — how big the new tissue grows — turned out not to match, and the same team did both the predicting and the measuring.

Technical: Rishabh R, Vučković J, Zadeh-Haghighi H, Beane WS, Simon C, PNAS Nexus 5(5):pgag152 (6 May 2026), doi:10.1093/pnasnexus/pgag152 — verified against the Oxford Academic record and the Zenodo code deposit. Constraint from the same cycle, preprint-tier and therefore carried as a caution rather than as evidence: Sung J, Antill LM, Cheong J, bioRxiv 10.64898/2026.07.15.738821, mapping spin-dependent radical-pair reactivity onto cryo-EM oxygen-accessible sites in mitochondrial complex I, finding the spin-sensitive hotspot ~3.3 Å from FMN collapses sharply with faster dephasing. The single-enzyme SOD…

Frozen prediction: By 2028-02-28, P=0.25 that an independent laboratory reproduces the non-monotonic superoxide-versus-field profile. Lowered from 0.30: planarian regeneration assays plus controlled sub-millitesla field exposure is a rare skill combination and no second lab is visibly set up for it.

DEMOTED. The measurement is real: cryptochrome 4a binds lipid bilayers together with the myristoylated alpha-subunit of the cone-specific G protein from European robin, with dissociation constants of 51-130 nM — affinities in the signalling range, not sticky-protein background. The proposer filed it at modeled on the grounds that 'SPR binding constants are replication-grade measurement'. That conflates the measurement CLASS with the FINDING. The finding is one laboratory (Koch, Oldenburg) extending its own prior claim — the same group published 'Direct Interaction of Avian Cryptochrome 4 with…

★ For the kids: Birds might sense magnetism with a protein in the eye, but nobody knew how that protein tells the brain. One lab found which protein it grabs onto and how tightly. The same lab said something similar before, and nobody has checked whether a magnet changes the grab at all.

Technical: Güzelsoy-Flügge Ü, Vujinović S, Scholten A, Koch KW, 'A supramolecular assembly of cone-specific G-protein and cryptochrome 4a on lipid bilayer', FEBS J 293(11):3362-3375 (2026), doi:10.1111/febs.70425 — verified against the Wiley record. Myristoylation of Gtα load-bearing for the interaction. Preprint-tier material struck from the evidence line: bioRxiv 10.64898/2026.06.29.735184 (robin retinal transducin profiling) and 10.64898/2026.05.19.726335 (AlphaFold3 LWO-Gtα model). The magnetic leg is entirely absent — every measurement to date is field-free — and that, not the tier, is the decisive…

Frozen prediction: By 2028-06-30, P=0.20 that a published experiment shows ErCry4a-Gtα complex formation, or downstream Gtα nucleotide exchange, is magnetic-field dependent. Unchanged from the proposers — this is the right number and the right gate.

DEMOTED for the same single-lineage rule applied to mScarlet3 and Cry4a-Gtα, and the demotion costs the finding nothing that matters. Two Lohmann-lab papers: a brief strong pulse of the kind that flips a magnetite dipole disrupts the loggerhead turtle's magnetic MAP under a conditioning assay built so the compass is not used; and a small weak magnet taped to the head — but not mid-body, not posterior — breaks the behaviour. That is the first positive anatomical localization of a magnetic sense in a vertebrate, and the apparatus is a magnet and a food-reward tank. It is also one laboratory, on…

★ For the kids: Nobody knows where in an animal the magnet-sensor actually is. Someone tried the simplest possible thing: stick a little magnet on the head, then the middle, then the tail, and see which one breaks the turtle's sense of place. It was the head. One lab, one kind of animal, so far.

Technical: Lim DS, Goforth KM, Lohmann CMF, Lohmann KJ, 'Evidence for cephalic magnetic map receptors in sea turtles', J Exp Biol 229(10):jeb252113 (15 May 2026), doi:10.1242/jeb.252113 — verified against The Company of Biologists record; weak magnet at head, mid-body and posterior against a conditioned magnetic-map/food association. Mackiewicz AG, Glazener AM, Goforth KM, Lim DS, Lohmann CMF, Lohmann KJ, J Exp Biol 228(22):jeb251243 (2025), doi:10.1242/jeb.251243. Caveats the value rests on: a localized magnet perturbs a REGION, so a negative placement rules out a receptor there only insofar as the per…

Frozen prediction: By 2028-06-30, P=0.50 that the placement assay is applied to a bird or an insect in a peer-reviewed paper. Conditional on a bird, P=0.35 that head placement disrupts the magnetic MAP while leaving light-dependent COMPASS orientation intact.

Held at open, with a correction the proposers missed. Three routes, one direction: nitric oxide synthase expression falls under both a strong field and removal of the geomagnetic horizontal component; RNAi knockdown of NOS downregulates Cryptochrome 2 and abolishes magnetic orientation; pharmacological inhibition with L-NAME reproduces the loss. What keeps it at open is a control not in the paper: NOS is pleiotropic in insect brains — learning, olfaction, general behavioural state — so 'no longer orients' is equally compatible with 'the magnetic pathway is broken' and 'the termite is impaired…

★ For the kids: Termites navigate using Earth's magnetism. Scientists switched off one brain protein two different ways and the termites got lost. But that protein does many jobs, so 'lost' might just mean 'unwell' — and the same lab previously said termites use tiny magnetic particles too.

Technical: Wang W, Gao Y, Long Q, Jiang Z, Huang Q, 'The role of neuronal protein NOS in regulation of magnetic orientation in the subterranean termite Odontotermes formosanus', Insect Mol Biol (2026), doi:10.1111/imb.70060 — verified against the Wiley record and PubMed 42533218. Exposure: 0.1 mT strong field and horizontal-component elimination. Single laboratory, no rescue arm, no non-magnetic behavioural battery reported (locomotion, phototaxis, trail-following), and the NOS→Cry2 link is transcript-level correlation rather than a demonstrated signalling dependency. Genuinely novel fork nobody has pos…

Frozen prediction: By 2028-06-30, P=0.40 that follow-up work shows NOS knockdown in this system impairs at least one non-magnetic behaviour. A published NO-donor rescue restoring magnetic orientation under sustained knockdown promotes this to modeled.

New component, created to hold a sub-claim that had been riding on enzyme-hydrogen-tunnelling's viable tier without its own evidence. Two proposers wanted this retired on the strength of Warshel's 2026 centennial review. Retirement is refused, and the refusal is the substantive skeptic ruling in this domain. Retirement in this corpus requires positive refutation or abandonment by the claim's own proponents. What exists instead is a decades-old, still-active theoretical dispute — Warshel and colleagues on one side arguing tunnelling is anticatalytic and unenhanced relative to the reference sol…

★ For the kids: Something happening is not the same as something helping. Enzymes definitely use the quantum shortcut; whether they were built to take advantage of it is an argument scientists have been having for twenty years, and nobody has done the experiment that would end it.

Technical: The unsettled question is a comparison, not a phenomenon: the correct baseline is enzyme-versus-matched-reference-solution-reaction, not enzyme-versus-no-catalyst, and nobody has run the matched comparison experimentally. Warshel's position (Zhang & Warshel, Protein Sci 35:e70683, 2026 — verified) is that on that comparison the quantum contribution does not carry the rate. That is an argument about how to count, made computationally, by a lineage that has held the position for two decades. Kill condition for the claim: a matched-reference experimental comparison with a null result. Promotion …

Frozen prediction: By 2028-12-31, P=0.70 that no experimental paper appears demonstrating enzyme-enhanced tunnelling against a properly matched solution reference reaction — i.e. the claim remains unproven rather than either proven or refuted.

DEMOTED. The paper is real and worth carrying: Warshel, whose computational work built the field's foundations, argues in a centennial review of the Schrödinger equation that tunnelling in enzymes is anticatalytic and that there is no consistent theoretical or experimental evidence enzymes enhance it relative to the matching solution reaction, and works a bird-navigation electron transfer as a case study concluding classical control of potential surfaces and energy-gap fluctuations dominates there too. The proposer filed it at modeled by citing the corpus's precedent of filing critique-class …

★ For the kids: A very famous chemist says the quantum shortcut inside enzymes doesn't make them faster and might even slow them down. He has said this for about twenty years, and other chemists disagree. It's an argument, not a measurement, so we file it as unsettled.

Technical: Zhang A, Warshel A, 'How important are quantum mechanical effects in controlling biological functions: Enzymes, electron transfer and bird navigation', Protein Sci 35:e70683 (2026), doi:10.1002/pro.70683 — verified; June 2026. The paper also treats the anticatalytic nature of low-barrier hydrogen bonds and argues classical control of electron-transfer efficiency generally. Promotion to modeled requires a second independent lineage reaching the same conclusion with an independent method, or an experimental matched-reference comparison supporting it.

Frozen prediction: Carried by enzyme-tunnelling-catalytic-exploitation.

DEMOTED, and the demotion is about relevance, not correctness. The photosynthetic-coherence debate has been stuck on an asymmetry: coherence is inferred from spectroscopic signatures while the thing anyone cares about — whether transport actually goes faster — is a different observable that depends on how the state was prepared. Liebert and Scholes supply the missing bridge as a bound rather than a story: a resource-theoretic impact functional quantifying the maximum change initial site-basis coherence can induce in any chosen readout given fixed reduced dynamics. Three useful outputs: an upp…

★ For the kids: People argue about whether plants use a quantum trick to move energy. Someone worked out the maths for the most useful question: how much could the trick help AT BEST? They only tried it on made-up toy examples so far, not on a real plant.

Technical: Liebert J, Scholes GD, 'Operational bounds and diagnostics for coherence in energy transfer', J Chem Phys 164(22):224113 (14 June 2026), doi:10.1063/5.0333456 (arXiv:2603.09748) — verified against AIP; Princeton. Companion: 'Operational impact of quantum resources in chemical dynamics', J Chem Phys 164(17):174115 (2026). Promotion to modeled requires application to FMO, PE545 or LH2 with measured Hamiltonian parameters and reported bounds. This is the tool the corpus's standing existence-vs-exploitation question has lacked, and it should be demanded of any future functional-coherence claim, t…

Frozen prediction: By 2028-06-30, P=0.60 that the impact-functional framework is applied to a real light-harvesting complex with measured Hamiltonian parameters. Conditional on that, P=0.65 that the bound on coherence-induced improvement in trapping efficiency comes in below 10% — consistent with 'coherence present, not functionally exploited'. A bound above 10% would be the first quantitative support for functional exploitation this corpus has ever been able to file, and must be logged as a surprise.

Held at open and explicitly cited as an instrument, not as evidence for magnetoreception. Avian magnetoreception has been stuck at behaviour-plus-chemistry partly for anatomical reasons: a robin's brain is too deep and too large for the optical methods that would let you watch a magnetic response happen. A honeybee's is not. The preparation puts a bee under a two-photon microscope with a calcium indicator retrogradely loaded into the anterior optic tubercle, a higher-order visual centre, inside a shielded three-axis Helmholtz coil that controls field strength and polarity independently of the…

★ For the kids: You can't easily watch a robin's brain while it senses magnetism — it's buried too deep. A bee's brain is small enough to look right into with a microscope, so scientists built a rig to watch bee brain cells while a magnetic field switches on and off. They haven't reported what they saw yet.

Technical: Oesterle AS, Kiriş A, Haase A, 'Experimental platform to study the neuronal mechanisms of magnetoreception in the honey bee', bioRxiv 10.64898/2026.05.05.722881, posted 8 May 2026 — verified; University of Trento. In vivo two-photon calcium imaging of the anterior optic tubercle via minimally invasive retrograde labelling; shielded three-axis coils. PREPRINT — cannot carry a tier above open. Critical design tension on publication: bees also carry abdominal magnetite, so a magnetic response in a visual neuropil does not by itself localize the receptor. The separating contrasts are light-depend…

Frozen prediction: By 2028-06-30, P=0.40 that a peer-reviewed paper from this platform reports magnetically modulated neural activity in a honeybee visual neuropil. Conditional on such a report, P=0.50 that the response is light-dependent (abolished in darkness) — the radical-pair signature — versus polarity-sensitive and light-independent, which points to abdominal magnetite and strengthens the corpus's classical reading.

Two proposers filed this twice (time-dependent-metrology-limits, driven-qubit-metrology-time-scaling); merged. Held at open: an unrefereed preprint with no experiment. Two corrections to how the proposers characterized it. First, the abstract's actual claim is more general and more conditional than 'T^4 under dephasing, T^3 under spontaneous emission': coherent dynamics giving T^(2k) scaling retain T^(2k) under parameter-independent noise in one regime and drop to T^(2k-1) in another, with the quoted exponents being the k=2 instances. Second, and more important for the corpus: the exponent co…

★ For the kids: A maths paper says a quantum sensor that you wiggle cleverly gets better and better the longer you watch, even with noise. It is pencil and paper so far. And the 'better than the usual limit' headline depends on what you count as the cost — time, or energy, or number of looks.

Technical: Previdi L, Albarelli F, 'Precision limits for time-dependent quantum metrology under Markovian noise', arXiv:2605.18392v1, submitted 18 May 2026 — verified via the arXiv API; no journal reference or DOI. Method: extension of the minimization-over-purifications framework to time-varying continuous channels, with a differential upper bound on quantum Fisher information via semidefinite programming; universal long-time scaling laws for parameter-independent noise, distinguishing DHNLS and DHLS regimes. Contrast with static-parameter results: Heisenberg T^2 collapses to T under generic Markovian …

Frozen prediction: By 2027-08-25, P=0.60 that the preprint appears in a refereed journal (promote to modeled). By 2027-12-31, P=0.12 that any experiment measures a precision scaling exponent >2 in total interrogation time for a time-dependent field with a noisy driven qubit (lowered from the proposers' 0.15: no group is visibly set up for the control sequence).

VERIFIED: Lisowski et al., Science 392:985-991 (28 May 2026, doi:10.1126/science.ady2486). 34 pigeons trained on a 19 km route; macrophage depletion (clodronate) abolished same-day homing under overcast (0 depleted birds home vs controls) while sunny-sky homing was unimpaired; vagal route proposed. Same-issue perspective (Spiro & Drakesmith, Science 392:919) hedges 'if replicated'. Held at open: single lineage, no rescue, transduction unmeasured, and clodronate is a systemic macrophage-depleting drug so off-target physiology is an open confound. Skeptic note: this causal-depletion result is S…

★ For the kids: Scientists found tiny magnetic immune cells in a pigeon's liver, and when they removed them the birds got lost on cloudy days. One team, one experiment — we wait for another team to check.

Technical: Depletion is causal but not causal-with-rescue; effect is cue-conditional (overcast). Open: how a liver-resident population encodes direction; whether ferritin superparamagnetism exceeds Brownian torque at 50 uT; two Science papers, two organs, zero convergence.

Frozen prediction: By 2028-09-05 no laboratory without shared authors with Lisowski et al. publishes a peer-reviewed replication of macrophage-depletion-induced loss of overcast homing (p=0.65).

VERIFIED: Lam & Malkemper, Physiol Rev 106:2383 (2026; doi:10.1152/physrev.00032.2025) state there is no evidence for magnetic compass orientation in squid, cuttlefish or octopus. Status is largely UNTESTED, not tested-null — the review is the sole anchor and it records an absence, so this component may not be read as evidence of absence.

★ For the kids: Octopuses and squid travel huge distances and have big brains, yet nobody has shown they feel Earth's magnetism. Maybe nobody looked properly.

Technical: Falsifier: two-alternative conditioned magnetic-direction assay in Sepia with coil rotation, blind scoring, n >= 40. Bridge to the octopus crew: a standing null, not a mechanism to propose.

Frozen prediction: By 2027-09-05 no peer-reviewed paper demonstrates magnetic compass orientation in any cephalopod (p=0.85).

VERIFIED: Russo, Allen, ... Winklhofer, Brady, Packer, Murray, Gilbert, Science Advances 12:eaed7391 (2026; preprint arXiv:2603.20312). Magnetometry on powdered dried specimens: 74/96 species show signal, present in outgroups, no phylogenetic signal, scales with body size and sociality. Abstract says 'putative magnetoreception'. A remanence screen on ground tissue cannot separate biogenic magnetite from ingested or adhering iron; a body-wide signal is the opposite of a localized receptor.

★ For the kids: Almost every kind of bee tested has a little magnetic material inside. Being magnetic is not the same as having a compass — iron dust would look the same.

Technical: What raises it: element-specific mapping (synchrotron XRF / TEM of single-domain crystals) in a defined tissue with clean-diet rearing in >= 2 species. Bee crew: log as a life-history correlation, not compass evidence.

Frozen prediction: By 2028-03-05 no peer-reviewed study localizes single-domain biogenic magnetite to a defined innervated tissue in a non-Apis bee reared on an iron-controlled diet (p=0.7).

Burd et al., Nature (2026) propose remote control of processes such as gene expression; only fluorescence modulation is shown. Same lineage as fluorescent-protein-giant-mfe — three components (this, the giant-MFE, the ODMR readout) all lean on the same Stanford paper and must not be summed as independent support. Kept distinct from magnetogenetics-controversy (ferritin-channel actuators).

★ For the kids: Scientists built a glowing protein into a worm and could turn its glow up and down with radio waves. Steering real worm biology with it is still a plan.

Technical: Falsifier for promotion: RF-gated transcription or enzymatic output in an animal, from a second group.

Frozen prediction: By 2027-09-05 no journal paper demonstrates RF-resonant control of gene expression or enzymatic activity (not fluorescence) in a living animal via an engineered radical pair (p=0.75).

VERIFIED: Cogitate Consortium (Ferrante et al.), Nature 642:133-142 (5 Jun 2025; doi:10.1038/s41586-025-08888-1): preregistered adversarial test of IIT vs GNWT, n=256, both theories lost preregistered predictions. Orch OR has entered no such protocol; its 2025-26 literature is proponent-authored. Skeptic ruling: the instrument exists as ONE consortium's single published application, and the component's gradable content is an absence claim about Orch OR — which sits in the awareness domain the constitution requires be open and flagged, never asserted. 'Modeled' is a mechanism-plausibility tier…

★ For the kids: Scientists who disagree about how the mind works agreed BEFORE the experiment what would prove each wrong. The 'quantum mind' idea has not joined that game.

Technical: Applicable to Orch OR: proponents would pre-sign which microtubule-drug x anaesthetic x species combinations MUST delay LORR and which MRI controls count as null. Bridge to the bee crew: the same instrument is the missing test for RPM-vs-magnetite.

Frozen prediction: By 2028-12-31 no preregistered, theory-neutral adversarial collaboration (OSF or equivalent) with Orch OR proponents as signatories of divergent predictions exists (p=0.85).

Threshold theorems assume independent errors; ionising radiation breaks Cooper pairs chip-wide. VERIFIED journal lineages: Google (McEwen, Nat Phys 18:107, 2022; PRL 133:240601, 2024 gap engineering; Kurilovich et al., 'Correlated Phase Error Bursts in a Gap-Engineered Superconducting Qubit Array', Phys Rev X 16:021025, 4 May 2026 = arXiv:2506.18228); MIT/Lincoln (Harrington, Nat Commun 16:6428, 2025; cosmic rays ~1/592 s, 17.1 +/- 1.3% of bursts); BAQIS/CAS (Li, Nat Commun 16:4677, 2025; muon vs gamma split on 63 qubits); Fermilab/Northwestern (Bratrud, Nat Commun 16:9906, 2025; 107 m underg…

★ For the kids: Particles from space and from the walls hit the quantum chip every few minutes and make many qubits fail at once. Four different labs have now proved it, and the fix only removes one kind of the damage.

Technical: QP bursts from phonon-mediated pair-breaking after MeV-scale deposits; T1 bursts fixed by asymmetric junction gaps; residual frequency shifts up to 3 MHz for ~1 ms cause correlated phase errors. Consequence: Lambda measured in burst-excluded windows is an upper bound; the radiation floor appears in the logical-error FLOOR, not the slope.

Frozen prediction: By 2028-09-05 a peer-reviewed superconducting surface-code memory at distance >= 9 names radiation-induced correlated bursts as the dominant residual logical-error contribution, or requires explicit burst mitigation to hold its Lambda (p=0.6); AND by 2027-09-05 at least one lab outside Google/MIT/BAQIS/Fermilab publishes correlated-burst detection on a >= 20-qubit superconducting array (p=0.7).

Pantry lead claim:kraken:48eeb9b1ea06 chased to its real source and VERIFIED: Perlin, Steinberg, Criger, arXiv:2607.19574 (Jul 2026). Adjacent (verified): Zhao et al., arXiv:2604.16209. Filed at open, not modeled: every number is a SIMULATED logical error rate under a circuit-level noise model, and no search-discovered code has been run on hardware. The new axis is code DISCOVERY by optimisation, whose failure mode is overfitting to the simulator's noise model — the failure class the corpus records for ML decoders. Skeptic note: a component spawned from preprints only sits at the floor and ma…

★ For the kids: Instead of a human designing the error-correcting recipe, a computer tries millions of recipes and keeps the best; nobody has cooked one on a real quantum machine.

Technical: Search objective = distance + minimum-weight logical-operator count; SWEL subset preserves transversal gates. Open question: whether syndrome-circuit depth and hook errors on real connectivity erase the simulated advantage.

Frozen prediction: By 2027-09-05 a hardware demonstration on >= 50 physical qubits of a code found by automated search reports logical error below the best physical error on the same device (p=0.35).

VERIFIED: Berut et al., Nature 483:187 (2012; colloidal bit); Yan et al., PRL 120:210601 (2018; single trapped-ion qubit, bound saturated); Aimet, Tajik, Tournaire et al., Nat Phys 21:1326 (2025; doi:10.1038/s41567-025-02930-9; 1D Bose-gas quantum field simulator). Skeptic caveat: Aimet et al. probe a GENERALISED Landauer relation (entropy change vs dissipated energy in a many-body quench), not a single-bit erasure — it supports the bound's generality, not a third erasure measurement. Two erasure platforms plus one many-body generalisation is still replication-grade for 'the bound holds'. Wat…

★ For the kids: Forgetting is not free: wiping a single yes/no from any memory, even a quantum one, warms the room by a tiny fixed amount. Three labs tried to beat it and could not.

Technical: Q >= kT ln2 per erased bit on MEAN heat; single-shot fluctuations below the bound are allowed and observed — the standard way a 'sub-Landauer' headline is manufactured.

Frozen prediction: By 2028-09-05 no peer-reviewed experiment reports MEAN erasure heat below kT ln2 for a bit erased with success >= 0.99 that survives 12 months without a published rebuttal (p=0.95).

KyberSlash (Bernstein et al., IACR TCHES 2025(2):209-234): secret-dependent division timings in official Kyber/ML-KEM reference code, keys recovered on Cortex-M4 and Raspberry Pi, patched after disclosure. Power/template attacks on pair-pointwise multiplication (IACR ePrint 2025/1577) and SPA-assisted chosen-ciphertext attacks (Springer CASCADE 2025). Multiple independent groups, reproducible artifacts = viable. Migration to ML-KEM moves the attack surface from number theory to constant-time engineering.

★ For the kids: The new codes are built to survive quantum computers, but the chips running them can still give secrets away by how long they take or how much power they use — and that has already happened.

Technical: Countermeasures: constant-time division rewrites, Barrett-style reductions; masking/shuffling cost 2-5x on embedded targets.

Frozen prediction: By 2027-09-05 at least one high-severity CVE is filed against an ML-KEM or ML-DSA implementation in OpenSSL, BoringSSL, liboqs, or a major HSM firmware whose root cause is implementation (timing/power/fault), not mathematics (p=0.7).

PHARAO cold-caesium clock + SHM maser + MWL/ELT links installed on Columbus, April 2025; mission papers (Cacciapuoti et al., arXiv:2411.02912) target a redshift measurement ~10x beyond the Galileo eccentric-orbit tests. As of Sept 2026 no primary science result has appeared. Skeptic clarification: the modeled tier attaches to 'hardware real and operating in orbit', not to any redshift number — the science claim is open until data. Same rule as the thorium clock: a launched instrument is not a result.

★ For the kids: A clock on the space station ticks a hair faster than clocks on the ground because gravity is weaker up there; the first numbers are still being counted.

Technical: Target fractional stability ~1e-16 at 10 days; goal ~2-3e-6 on the redshift parameter vs 2.5e-5 from Galileo (Delva et al. PRL 2018; Herrmann et al. PRL 2018).

Frozen prediction: By 2028-04-30 an ACES gravitational-redshift result is published in a peer-reviewed journal with fractional uncertainty <= 3e-6 (p=0.5).

Pileio, Carravetta & Levitt, PNAS 107:17135 (2010; low-field singlet storage); Stevanato et al., Angew Chem Int Ed 54:3740 (2015; 13C2-naphthalene singlet lifetime > 1 h at room temperature, ~50x T1). The phenomenon (minutes-scale singlet order in solution) is replicated across groups (Southampton, ENS Paris, Duke) and nuclei (13C, 15N, 1H) = viable; the >1 h record itself is Levitt-group. It is nuclear, not electronic, and requires symmetry-engineered spin pairs. Anchor for what the Posner/Fisher hypothesis borrows: a 31P singlet lifetime in a Posner cluster has never been measured (Swift, V…

★ For the kids: Tiny magnets inside atomic nuclei can keep a quantum pattern for over an hour in a warm liquid — but only in molecules chemists built on purpose. Nobody has measured it in anything living.

Technical: Singlet order is immune to intra-pair dipolar relaxation; lifetime needs near-equivalent pairs and suppression of CSA relaxation.

Frozen prediction: By 2027-09-05 no peer-reviewed paper reports a measured 31P long-lived-state lifetime in a Posner cluster, synthesized or biological (p=0.85).

Dissolution-DNP [1-13C]pyruvate MRI imaged in humans since Nelson et al., Sci Transl Med 5:198ra108 (2013), across prostate, brain, breast, liver, kidney at multiple centres (Wang et al., Radiology 291:273, 2019). The field's own literature does not brand it quantum — the hype-boundary datum. Status: investigational (IND only), no NDA/BLA. The 'Commun. Chem. 2026 ultra-low-field' citation was not verified by the skeptic and is downgraded to a watch-item. NO claim that coherence (vs polarization) is exploited.

★ For the kids: Doctors can make a sugar molecule glow inside an MRI by lining up its atomic spins in a super-cold machine. It works in real patients — and nobody needs to call it 'quantum'.

Technical: Polarization (population imbalance) != coherence. Bottleneck is polarizer capex and ~1-min T1 logistics, not physics.

Frozen prediction: No FDA NDA/BLA approval of hyperpolarized [1-13C]pyruvate as a diagnostic agent before 2027-12-31 (p=0.85).

VERIFIED: Muradoglu et al. (Q-CTRL), arXiv:2504.08167 (10 Apr 2025), no journal-ref: airborne trials to ~19,000 ft, 'up to 46x' and 'at least 11x' better positioning than a velocity-aided strategic-grade INS; ground 7x. (The '94x' figure in press coverage is not in the abstract.) 17 months as a vendor preprint; the July 2025 Australian Defence sea trial is a vendor blog. NEW, verified: Scholes et al., arXiv:2606.24612 (23 Jun 2026; Strathclyde-lineage authors, affiliations not on the abstract page) — wearable FID-OPM MagNav, 2.24 m radial error over a >500 m route, baseline is dead-reckoning,…

★ For the kids: Instead of GPS, a super-sensitive quantum compass reads Earth's magnetic fingerprint and matches it to a map — like turtles are thought to do. One company says it works well; a university group has tried a wearable version; nobody has checked the company's numbers.

Technical: Bridge: map-matching (position) vs compass (heading) are different computations; the bee evidence is compass-only.

Frozen prediction: By 2027-09-05 no independent (non-Q-CTRL) peer-reviewed field trial reports quantum-magnetometer MagNav beating a strategic-grade INS (p=0.75).