What was reported, what the source supports and what remains unestablished. Read the conditions alongside the claim.
18 source-checked records · A bounded reading collection. Source checks are not independent replication or scientific certification.
A source is authoritative for a specific question
This six-record reference sample keeps the original ledger claims and adds exact locations, source roles and citations. The new locator checks do not reset the historical claim-review dates.
A source family groups the same underlying announcement or research record. Several links do not necessarily provide independent confirmation.
ROADMAP WATCH · FIRST EDITION
Three commitments worth following.
What is promised, when the source says to expect it and what evidence would establish delivery. This selected snapshot was reviewed on . Every item below remains a target in this edition.
IonQTarget · Late 2027
Planned IonQ system arrival at FIU
FIU says the IonQ system is expected to arrive in late 2027.
Not established
The announced arrival window does not establish completed delivery, a contract price or a revenue-recognition schedule.
Evidence to look for next
Has the planned campus installation been delivered and made accessible to researchers?
Original wording and source trail
The university announcement records a planned campus arrival. The existing Market Desk milestone remains the source of this record.Read that source
IBM targets a Starling system running 100 million quantum gates on 200 logical qubits in 2029.
Not established
These are planned system capabilities. A component prototype or result on another IBM processor does not establish delivery of Starling.
Evidence to look for next
A named Starling system result reporting the logical resources, executed circuit size, error conditions and access arrangements together.
Original wording and source trail
The June 2025 announcement specifies 2029 and the logical-qubit and gate targets.Read that source
The hardware page inspected for this edition still describes Starling as planned for 2029.Read that source
IBM: path to a fault-tolerant quantum computerOriginal company targetSource published · Recorded source check Find: Opening paragraph and The IBM roadmap to fault-tolerant quantum computing
IBM hardware and roadmapCurrent company roadmap pagePublication date not displayed by the source · Recorded source check Find: IBM Quantum Starling section
Record ibm-starling-watch-r1 · Snapshot reviewed 2026-10-15. A retrieval is not a new company announcement.
Quantinuum’s second-quarter 2026 updateLater company schedule statementSource published · Recorded source check Find: Product Technology and Supply Chain: Apollo schedule bullet
Record quantinuum-apollo-watch-r1 · Snapshot reviewed 2026-10-15. A retrieval is not a new company announcement.
Year and window labels retain the source’s precision. An elapsed target prompts another review; it does not prove success or failure. Technical results and financial outcomes require separate evidence.
QUBITWIRE RESEARCH NOTE · 15 OCTOBER 2026
A quantum roadmap is only useful if you can tell what would count as delivery.
A campus installation, a fault-tolerant system and a published schedule answer different questions. Reading them as equivalent milestones obscures what still needs to happen.
The three records in this edition were selected to illustrate those differences. They cover named targets from IonQ’s university partner, IBM and Quantinuum. The sample is deliberately small and is not a census of quantum computing companies, a completion-rate study or a ranking.
Separate the promise from the evidence.
For FIU’s planned IonQ installation, the next useful evidence is delivery and researcher access at the named campus. A schedule does not establish either. For IBM Starling, the stated logical resources and circuit size need to be demonstrated together under documented conditions. Neither item can be settled by counting a different processor’s physical qubits.
Quantinuum’s Apollo record illustrates why the dated trail matters. The 2024 announcement uses a by-2030 formulation; the 2026 update specifies 2029. Keeping both lets readers inspect the evidence without presenting the later wording as if it had always been the original statement. These formulations alone do not establish a delay or a delivered system.
Use three separate checkpoints.
Commitment. Identify the named machine or installation, the promised outcome and the precision of the target window.
Demonstration. Look for results from that same system, with the task, error conditions and resources stated together.
Access. Check who can use the delivered capability and under which practical conditions. A laboratory result does not automatically establish an operational service.
Return when the evidence changes.
A new technical result, an explicit schedule revision or documented system access can change the assessment. A refreshed web page or an elapsed year is not enough by itself. These are research questions, not assumptions that any company will succeed or fail.
Method, scope and reuse
Edition roadmap-watch-2026-10-15 uses the three source-linked records above, observed by 15 October 2026. Original publication dates range from September 2024 to September 2026; the current IBM hardware page does not display a publication date. Source dates, recorded checks and this edition’s review date are kept separate.
The downloadable JSON contains the same record IDs, target precision, source roles, limitations and source-dated history shown here. It includes no private notes, quote data or completion scores. The IonQ installation retains its existing Market Desk milestone ID. Other technical evidence links provide context and do not close these targets.
QubitWire’s interpretation is the reading framework above. The underlying schedules are attributed company or university statements. Corrections should identify the record and the supporting source. Submit a correction.
Caune and colleagues report mean decoding below one microsecond per round in an eight-qubit stability experiment.
Supported observation
The measured mean ranged from 0.44 to 0.79 microseconds over 5–25 decoding rounds. A separate feedback experiment measured 9.6 microseconds for nine measurement rounds, including communication and control.
Conditions
A Collision Clustering FPGA decoder operated at 156.25 MHz within Rigetti Ankaa-2. The stability-8 circuit omitted the central X stabilizer. The per-round figure is a mean; some runs exceeded one microsecond.
Not established
This small experiment does not establish sustained streaming performance for larger codes. Mean decoding time is not a worst-case guarantee or full feedback latency.
HRL's team reports a silicon quantum processor connected to a 4-kelvin CMOS controller through a superconducting ribbon cable.
Supported observation
The controller generates time-varying signals for a millikelvin chip with 54 quantum dots, configurable as up to 18 exchange-only qubits. The study includes repetition-code and quantum error-detection experiments.
Conditions
The integrated prototype retains room-temperature digital communication, static biases and readout digitization. Fifty-four dots is not fifty-four qubits. The error-detection demonstration used six physical qubits and postselection.
Not established
The prototype does not establish a utility-scale fault-tolerant computer. The paper identifies remaining interconnect, controller-power, device-uniformity and system-architecture work.
Review: Source and editorial check, 2026-09-26. Attributed primary-source summary; no independent scientific review. Independent replication: not assessed.
Record first published 2026-09-26 · Revision claim-hrl-cryogenic-control-integration-r1.
Inspect the exact evidence and citation
This source-locator check adds context to the existing record. Its original publication and claim-review dates remain unchanged. No independent scientific replication was performed.
Evidence context
Reported hardware experiment
Independent replication
not assessed. This describes what was assessed, rather than evidence that replication failed.
Members of the HRL Quantum Team and Collaborators. A digitally controlled silicon quantum processing unit.
Nature. Source published 2026-07-29. Published journal article, 29 July 2026.
Abstract; Main, system architecture; Figure 1. https://www.nature.com/articles/s41586-026-10754-7#Fig1
Source authority: The reported 4 K controller, ribbon interconnect and configurable quantum-dot system.
Does not establish: Independent replication or a complete useful fault-tolerant application.
QubitWire: Cryogenic control connects to a silicon processor. Record first published 2026-09-26. Cited revision claim-hrl-cryogenic-control-integration-r1.
https://qubitwire.com/evidence/hrl-cryogenic-control-integration?revision=claim-hrl-cryogenic-control-integration-r1
Current record and corrections: https://qubitwire.com/evidence/hrl-cryogenic-control-integration
Limitation: The prototype does not establish a utility-scale fault-tolerant computer. The paper identifies remaining interconnect, controller-power, device-uniformity and system-architecture work.
Atom Computing and collaborators report toric-code memory experiments with replacement atoms and repeated reservoir reloading.
Supported observation
The smaller code ran for up to 90 syndrome-extraction cycles; the larger code reached 70. Storage-zone atoms were replenished every tenth cycle.
Conditions
The preprint compares codes using 16 data plus 16 ancilla qubits and 32 data plus 32 ancillas. Recorded measurements were decoded with PyMatching using loss-aware reweighting. Long runs and the short-run code-size comparison are separate datasets.
Not established
The 90-cycle result does not demonstrate indefinitely accurate computation or universal fault tolerance. Lower error for the larger code in short runs must not be generalized to the deep-run results.
IBM reports initial testing of its modular cryogenic prototype, with two cells coupled and operating together.
Supported observation
The technical information sheet documents prototype cooling to 4 kelvin and base temperature. Rectangular cells provide flat interfaces for connecting adjacent cooling environments.
Conditions
The inspected information sheet is version 2.1, dated 18 August 2026; its repository entry was published the next day. IBM’s accompanying report identifies the two-cell demonstration in Poughkeepsie. IBM describes a change from earlier cylindrical cooling enclosures; no computational comparison is established here.
Not established
Cooling connected cells does not demonstrate fault-tolerant computation across chips. Deployment timing and future qubit capacities are plans, not measured computing outcomes.
D-Wave’s research team reports an approximately 500-nanosecond controlled-Z gate between two dual-rail cavity qubits.
Supported observation
The Nature paper reports about 0.5% erasures per gate and residual Pauli errors below 0.1% in its characterization.
Conditions
Two encoded qubits use superconducting microwave cavities and a transmon coupler. Residual errors are evaluated after detected erasures are removed; they must be read alongside the erasure rate. Characterization uses state tomography and interleaved randomized benchmarking. Larger-code comparisons are simulated.
Not established
This is not a demonstrated large error-corrected processor. Error behavior worsens at greater repeated-gate depth; the surface-code simulation assumes perfect erasure checks and restricted noise models.
Sivak and colleagues report reinforcement-learning control that reduces variation in logical error rates under deliberately injected drift.
Supported observation
Control steering improves stability 2.4-fold against a fixed policy; adding decoder steering raises this to 3.5-fold. Stability here refers to the standard deviation of logical error rates.
Conditions
The hardware experiment uses short quantum-memory runs on Willow with states prepared again for each shot. The decoder-steering method also estimates logical errors. The drift comparison uses control parameters fixed at their initial calibration.
Not established
This does not demonstrate uninterrupted execution of a long algorithm. The reported decoder steering has real-time scaling limitations; large-code scaling and continuous steering are studied in simulation.
Pasqal announced that a photonic chip delivered four traps holding four rubidium atoms, with an approximately 27.5-second trapping lifetime.
Supported observation
The inspected primary announcement contains those figures. This source check confirms what the company announced; the underlying experimental methods and data were not independently inspected.
Conditions
The announcement credits work with Aeponyx and describes a four-atom demonstration. Pasqal compares trapping performance with its existing bulk optics, without providing a detailed comparison dataset in the announcement.
Not established
An announcement does not verify gate fidelity, reproducibility or a larger processor. Projected footprint reductions and future atom counts are not demonstrated results in this record.
Canada announced C$195 million for Xanadu’s C$893 million research and manufacturing project.
Supported observation
The August 28 announcement identifies federal support through the Strategic Response Fund for expanded research facilities and advanced manufacturing in Canada.
Conditions
Planned capabilities include packaging, testing, assembly and integration of photonic and semiconductor components for quantum systems. These are announced funding and project amounts, not measured quantum-computer performance.
Not established
The announcement does not establish completed facilities, industrial production or demonstrated fault-tolerant computing. Project delivery remains prospective.
RIKEN reports that its ROQUO GPU supercomputer has entered operation in Kobe.
Supported observation
The launch report describes 135 compute nodes with 540 GPUs and an SQC interface for hybrid system development.
Conditions
The June 19 report identifies integration with Fugaku, ibm_kobe and Reimei as an anticipated research outcome. External allocation is planned through a test-user program. The reported 19.80 PFLOPS HPL result measures classical FP64 computation across ROQUO, not quantum-HPC application speedup.
Not established
An operational GPU platform does not establish end-to-end hybrid advantage or unrestricted public access. This record does not verify current admission availability.
Lo and colleagues report a universal topological gate set using braiding and fusion on Quantinuum H2.
Supported observation
The checked v2 preprint prepares a 54-qubit S3 topologically ordered state, encodes logical qutrits and demonstrates magic-state preparation.
Conditions
The experiment combines braiding with charge measurements. Magic-state preparation postselects specified fusion outcomes, so accepted shots are conditional results. The record concerns demonstrated operations within a programmed trapped-ion system, not a comparison with commercial fault-tolerant computers.
Not established
This proof of principle does not establish scalable fault tolerance or an application advantage. The journal version was inaccessible; this assessment is limited to the inspected preprint.
IBM’s QOBLIB update describes a shared library of ten optimization problem classes for comparing quantum and classical methods.
Supported observation
The August 12 update reports public problem instances, evaluation metrics, classical baselines and a website recording best-known results.
Conditions
Results apply to specified problem instances and solver settings. The announcement describes a changing community repository, not a frozen performance ranking. Useful comparisons require the same task, solution-quality target and resource accounting against strong current classical methods.
Not established
A benchmark library is not evidence that quantum optimization has achieved practical advantage. This source check does not audit individual submissions or certify their reproducibility.
Review: Source and editorial check, 2026-09-26. Attributed primary-source summary; no independent scientific review. Independent replication: not assessed.
Record first published 2026-09-26 · Revision claim-qoblib-shared-benchmark-resource-r1.
Inspect the exact evidence and citation
This source-locator check adds context to the existing record. Its original publication and claim-review dates remain unchanged. No independent scientific replication was performed.
Evidence context
Benchmark resource; no new experiment performed here
Independent replication
not assessed. This describes what was assessed, rather than evidence that replication failed.
BREAKTHROUGH LEDGER · classical preprocessing study
Acharya and colleagues report portfolio decomposition that reduces subproblem size by about 80% in their tested data.
Supported observation
For linear-constraint problems around 1,500 assets, the study reports roughly threefold runtime savings with solution quality within 5% of the optimum.
Conditions
The checked v2 preprint reports Gurobi benchmarks, sequential subproblem solving and explicit approximation gaps. Its decomposition uses correlation preprocessing, clustering and risk rebalancing. Timing compares decomposed and direct optimization using the same classical solver. It is not a quantum-versus-classical timing result.
Not established
Results depend on the tested data and constraints. Potential compatibility with smaller quantum devices does not demonstrate quantum speedup or production portfolio returns.
DARPA selected Atom Computing, Diraq, IBM and IonQ for QBI Stage C on October 7.
Supported observation
Stage C moves the four approaches into government verification and validation. Microsoft and PsiQuantum had already entered through the precursor programme.
Conditions
The status concerns a proposed route to utility scale. Company releases are interpretations of the same programme event, not additional independent tests.
Not established
Selection does not establish a successful useful application, completed fault-tolerant system or technical winner.
Four more teams enter QBI’s final stageDARPA · Government programme announcement · Source published Source checked Source family: qbi-stage-c-2026-10-07. Programme status originator
IBM advances to QBI Stage CIBM · Company announcement · Source published Source checked Source family: qbi-stage-c-2026-10-07. Participating company interpretation
IonQ advances to the final QBI stageIonQ · Issuer announcement · Source published Source checked Source family: qbi-stage-c-2026-10-07. Participating company funding disclosure
No automatic score change; the published QRank methodology and current edition remain authoritative.
Review: Source and editorial check, 2026-10-08. Primary-source wording and scope review; no independent scientific reproduction. Independent replication: not assessed.
The QUOPS v1 preprint proposes a cross-platform benchmark of executable circuit size and execution speed.
Supported observation
Its abstract reports physical-qubit experiments on Quantinuum, Google and IBM; a separate encoded-logical-processor assessment; and future fault-tolerant projections.
Conditions
These are three distinct scopes. Challenge-problem mappings depend on their resource assumptions. This record reviews the abstract and version identity; individual table values are not transcribed.
Not established
No independently reproduced benchmark, universal platform winner or replacement QRank metric is established here.
No automatic score change; the published QRank methodology and current edition remain authoritative.
Review: Source and editorial check, 2026-10-08. Primary-source wording and scope review; no independent scientific reproduction. Independent replication: not assessed.
Record first published 2026-10-08 · Revision claim-quops-method-2609-12146-v1-r1.
Inspect the exact evidence and citation
This source-locator check adds context to the existing record. Its original publication and claim-review dates remain unchanged. No independent scientific replication was performed.
Evidence context
Reported physical experiments, logical assessment and projections kept separate
The partners say earlier work trapped a two-dimensional array of nine individually controlled ions.
Supported observation
Their October 7 announcement expands a collaboration on trapped-ion architecture, semiconductor manufacturing, packaging and photonics.
Conditions
October 7 is the announcement date, not a verified experiment date. The two company pages describe one announcement family; the array result remains attributed to the partners.
Not established
A direct methods review, independent replication, manufacturing yield, complete multi-qubit performance or an error-corrected application is not established by these releases.
Infineon and ZuriQ deepen their collaborationZuriQ · Joint company announcement · Source published Source checked Source family: infineon-zuriq-2026-10-07. Partner account of the same announcement
Infineon and ZuriQ deepen their collaborationInfineon · Joint company announcement · Source published Source checked Source family: infineon-zuriq-2026-10-07. Issuer account of the same announcement
No automatic score change; the published QRank methodology and current edition remain authoritative.
Review: Source and editorial check, 2026-10-08. Primary-source wording and scope review; no independent scientific reproduction. Independent replication: not assessed.
Record first published 2026-10-08 · Revision claim-zuriq-array-prior-demonstration-r1.
BREAKTHROUGH LEDGER · software release and integration
NVIDIA announced CUDA-Q Logical as an orchestration layer for fault-tolerant system development on September 14.
Supported observation
The release describes modelling, design workflows and QUOPS integration. The official CUDA Quantum repository provides the implementation source.
Conditions
Development time, modelled resource use and physical application runtime are separate quantities. No repository version was built or tested during this source review.
Not established
The software announcement does not demonstrate that a proposed fault-tolerant application has executed on completed hardware.
No automatic score change; the published QRank methodology and current edition remain authoritative.
Review: Source and editorial check, 2026-10-08. Primary-source wording and scope review; no independent scientific reproduction. Independent replication: not assessed.
Record first published 2026-10-08 · Revision claim-cuda-q-logical-2026-09-14-r1.
The March 30 v1 preprint estimates cryptographically relevant Shor tasks using reconfigurable atomic qubits.
Supported observation
The abstract describes a physical-resource and runtime trade-off based on error-correcting codes, logical instructions and circuit design.
Conditions
Factoring and discrete logarithms are different tasks. Architecture, operating assumptions and parallelism determine the estimate. Precise error-rate inputs and full numeric comparisons were not reviewed in this record.
Not established
The theoretical analysis is not a built processor, completed cryptographic attack or confirmation that every neutral-atom architecture meets the assumed conditions. Later financing does not validate it.
Shor’s algorithm with reconfigurable atomic qubits · v1Cain and colleagues · arXiv · Theoretical research preprint · Source published Source checked Source family: arxiv-2603.28627. Research originator; abstract and version record reviewed
No automatic score change; the published QRank methodology and current edition remain authoritative.
Review: Source and editorial check, 2026-10-08. Primary-source wording and scope review; no independent scientific reproduction. Independent replication: not assessed.
Record first published 2026-10-08 · Revision claim-atomic-resource-estimate-2603-28627-v1-r1.
Quantinuum and UWA announced a non-binding MOU on October 5.
Supported observation
It proposes cloud access to Quantinuum’s platform, training and exploration of hybrid applications. A university-wide applications centre is exploratory.
Conditions
Definitive agreements and the planned activities may not follow. Access confirmation, an application result and financial disclosure would be distinct later events.
Not established
The MOU does not establish delivered access, a measured application benefit, a contract price or recognized revenue.
Quantinuum and UWA sign a quantum-access MOUQuantinuum · Issuer announcement · Source published Source checked Source family: quantinuum-uwa-2026-10-05. Agreement announcement originator
No automatic score change; the published QRank methodology and current edition remain authoritative.
Review: Source and editorial check, 2026-10-08. Primary-source wording and scope review; no independent scientific reproduction. Independent replication: not assessed.
Record first published 2026-10-08 · Revision claim-quantinuum-uwa-mou-2026-10-05-r1.
These reading notes explain the scope of recent reporting. They do not revise QRank scores, assessment dates or existing claim revisions. A source check is not independent replication.
A collaboration sets an agenda. What would show a result?
Quandela announced a biotechnology and drug-discovery collaboration with KRIBB on 7 October. The memorandum was signed on 8 September.
Source checked · Research contextConditions, limits and sources
Conditions: The announcement describes a research programme and planned cooperation, rather than a completed application test.
What remains unestablished: It establishes an agenda, not a clinical result, measured hardware improvement, deployed service or quantum advantage.
What to follow: Look for a defined task, methods, measured output and a published comparison. A plan to publish is not a paper.
One company announcement describes the agreement. It does not independently validate an application outcome.
The experiment encoded a small Z₂ lattice gauge theory with two qubits representing gauge fields and two vibrational oscillator modes representing matter. Changing the encoded flux changed the observed tunnelling through interference.
Source checked · Research contextConditions, limits and sources
Conditions: The physical resources, encoded model and measured observable are different parts of the claim. The loop combined digital preparation and readout with analogue evolution.
What remains unestablished: This is a specific simulation of dynamical gauge–matter interactions. It does not establish a universal computational advantage or a company-ranking result.
What to follow: Follow larger geometries, controlled errors and comparisons matched to the same model and observable.
The university report, paper and deposited data describe one experiment. Supporting data enable scrutiny; their availability is not evidence of independent reproduction.
Power and computational performance measure different things
A laboratory prototype absorbed light, retained energy and delivered electrical current. The paper studies electrical-power scaling in a collective light–matter system.
Source checked · Research contextConditions, limits and sources
Conditions: Charging power, stored energy, electrical output and retention time need separate measures. None is a processor error rate or computational speed.
What remains unestablished: The proof of concept does not establish a consumer battery, practical vehicle charging or a computational-performance ranking.
What to follow: Look for usable capacity, conversion efficiency, longer retention and repeatable operation at a relevant scale.
The institutional release and paper refer to the same prototype, already covered in the existing QubitWire battery story.
A field trial needs a claim shaped like the field trial
A quantum gravimeter aided inertial navigation on an 83 km maritime route. The version 1 preprint reports about one nautical mile error over the first 70 km and 2.2 nautical miles at the endpoint.
Source checked · Research contextConditions, limits and sources
Conditions: The navigation chain excluded GNSS inputs and matched gravity measurements against a stored map. GNSS was used later to evaluate position error. A separate surveying mode used GNSS.
What remains unestablished: Company-led trial and preprint evidence does not establish independent replication, certification or reliable operation on every route. Sensor performance and navigation-system accuracy differ.
What to follow: Follow independent evaluation across routes, sea conditions, map quality and clearly matched baselines.
The company announcement and preprint describe one trial family. Surveying and navigation are distinct analyses, not independent replications.
Quantum-relevant security still runs on ordinary systems
Post-quantum cryptography uses conventional computers. Its connection to quantum computing is the threat it is designed to resist. Read availability, authentication and customer adoption as separate questions.
Cloudflare added ML-DSA support to Custom Origin Trust Store and the per-zone and per-hostname configurations of Authenticated Origin Pulls for Cloudflare-to-origin authentication.
Akamai identifies end-to-end post-quantum key exchange as generally available on its Enhanced TLS network from 12 August. Its origin behaviour remains disabled by default in the documentation checked here.
Source checked · Research contextConditions, limits and sources
Cloudflare origin authentication
Conditions: Custom Origin Trust Store requires Advanced Certificate Manager. Customers must configure compatible origin systems; global Authenticated Origin Pulls support was deferred in the announcement.
What remains unestablished: Support does not establish that every customer or visitor connection uses post-quantum authentication. The 2029 suite-wide objective remains a roadmap target.
What to follow: Follow documented enablement, deployment scope and disclosed customer use. Keep key exchange, authentication and adoption separate.
The technical release and earlier roadmap are provider-authored. A target date is not a delivered capability.
Conditions: The origin server controls the negotiated key exchange. Unsupported origins can request a classical alternative; already-open persistent connections retain their existing negotiation.
What remains unestablished: Availability is not universal enablement or complete post-quantum authentication. No adoption rate or incremental revenue is established.
What to follow: Follow service scope, explicit configuration, negotiated connections and disclosed adoption. Missing adoption information is unknown, not zero.
Reading the evidence
What makes a component result useful?
These notes connect dated component results, manufacturing plans, financing and software access. They leave QRank scores, assessment dates and existing claim revisions unchanged. A source check is not independent replication.
A fast link still needs a quality label
The version 1 preprint reports an averaged heralded Bell-pair rate of 1.03(1) kHz and state fidelity of 87.9(7)% between a trapped barium ion and a silicon-vacancy centre in diamond, including the photon-conversion loss.
Source checked · Research contextConditions, limits and sources
Laboratory experiment / preprint · Event dated
Different measures from one experimental protocol
Measure
Reported value
What it describes
Heralded Bell-pair rate
1.03(1) kHz
Average successful entanglement events per unit time in the stated protocol.
Entangled-state fidelity
87.9(7)%
Quality of the resulting state under the same experimental protocol.
Heralded ion–diamond Bell-pair generation, including conversion loss; version 1 snapshot.
Conditions: Rate and state fidelity describe different parts of the same experimental protocol. A successful event is a heralded entangled pair, not an application step. Parenthesized uncertainty notation is retained from the paper.
What remains unestablished: This component-level result is not deployed distributed-application throughput, revenue evidence or a universal best-link comparison. A separate local nuclear gate fidelity is not the reported link fidelity.
What to follow: Follow versioned journal evidence and integrated operations that count memory waiting, other links and complete operation costs under matched event definitions.
The preprint and IonQ announcement describe one experimental source family. The company release is not independent reproduction.
The research uses a deliberate protocol that competes scrambling dynamics with measurement and reset-based control toward an ordered fixed point. Its version 1 manuscript first appeared on 22 September 2025; the journal and Rutgers report are dated 9 October 2026.
Source checked · Research contextConditions, limits and sources
Deliberate circuit experiment · Event dated
Conditions: The task is to control specified dynamics. Ask which target state and observable define success, and which information the procedure was required to preserve.
What remains unestablished: An ordered final state does not establish protection of arbitrary unknown quantum information, diagnosis of every hardware error or a universal fault-tolerance threshold.
What to follow: Follow a precisely stated preservation task, expanded controlled experiment or a new manuscript version; retain the original preprint chronology.
The university report, earlier manuscript and journal publication describe the same research family. Collaboration credits do not automatically assign ownership or equal company scores.
The journal identity and university report were checked. Full publisher text was not retrieved; the version 1 author manuscript supplies the protocol detail.
A financing event cannot stand in for a link measurement
University of Sussex and participating investor Nauta Capital reported more than US$100 million in Series A financing for Universal Quantum’s modular trapped-ion programme.
The February 2023 paper studied transport of ion qubits between adjacent ion-trap microchip modules, with ion loss and phase coherence evaluated separately.
Source checked · Research contextConditions, limits and sources
Universal Quantum Series A
Private-company financing · Event dated
Conditions: The more-than amount and USD currency belong to financing. Intended development and commercialization are future uses of capital, not completed-system revenue.
What remains unestablished: The round does not establish valuation, public investment access, hardware performance or the quality of an inter-module link.
What to follow: Follow separately sourced technical or commercial outcomes with explicit dates and scope, rather than converting investor enthusiasm into capability scores.
The university spin-out report and participating-investor account concern the same financing. Both have a relationship to the programme; neither independently validates the machine.
Conditions: This matter-link physically moves an ion. Ion-transport rate and loss are not interchangeable with heralded Bell-pair rate, remote-gate fidelity or finished application throughput.
What remains unestablished: Historical transport evidence does not by itself establish a completed larger modular computer or a new October 2026 technical result. Financing cannot refresh its measurement date.
What to follow: Look for links and local operations working together, with timing, scheduling and repeated-use reliability included.
Qualification is the manufacturing milestone to watch
The multi-year partnership targets industrialization of superconducting nanowire single-photon detectors and ultra-low-loss silicon-nitride photonics on a 300 mm manufacturing platform in Malta, New York.
Source checked · Research contextConditions, limits and sources
Conditions: Wafer size, detector efficiency, yield and integrated-system performance are different quantities. NIST explains detector physics and tradeoffs; it supplies no measurements of the partners’ process.
What remains unestablished: The announcement does not establish a qualified production process, measured repeatability, shipments, revenue or a finished fault-tolerant computer. Missing yield is unknown rather than zero.
What to follow: Look for dated process qualification, defined yield, repeatability or integrated-component results. An evidence ladder is a reporting taxonomy, not a probability forecast.
The issuer-distributed release is the same event already covered in the canonical manufacturing article. NIST is general technical context, not validation of the manufacturing plan.
The original investor-relations page returned an access block. This is the issuer’s same release distributed through GlobeNewswire, not an independent source family.
D-Wave announced a selected-participant beta for a gate-model simulator focused on error-aware programming, before planned general availability and future physical gate-model systems.
The August paper reported a physical entangling gate for dual-rail erasure qubits encoded in pairs of superconducting microwave cavities, studying whether useful error structure survives the operation.
Source checked · Research contextConditions, limits and sources
Conditions: Keep the simulator’s model, version, included operations and access eligibility attached to any result. Named participation is distinct from paid use.
What remains unestablished: Software access does not establish measured physical gate performance, public access for every visitor, deployed gate-model hardware or undisclosed customer revenue.
What to follow: Follow a documented version or access change, versioned model assumptions and a measured comparison with the corresponding hardware.
The company announcement is one software-access event already covered by the canonical simulator story, not a new physical experiment.
Conditions: An erasure location can help a suitable code or decoder. Recovery still requires structured encoding; detector mistakes and unflagged errors must be counted. The paper’s larger correction comparisons include simulations.
What remains unestablished: The gate experiment is separate from October beta access. It is not a complete fault-tolerant computer, proof of perfect error flags or a universal reduction of every correction cost.
What to follow: Follow broader operations, measured error-model correspondence and full resource comparisons including imperfect detection.
The publisher record and official PubMed author abstract describe one paper. A repository or index copy is not independent reproduction.
A guide to the evidence
Five questions that make a quantum claim clearer.
Use these reading habits with a press release, a paper or a company profile. Each opens a worked example from the same shared evidence records.
Identify the claim before judging it
An announcement describes what an organization says or plans. Theory asks what follows under stated assumptions. A simulation tests a model. A measurement reports an observation under particular conditions. An interpretation explains what those results might mean.
These are different kinds of work, not a ladder from bad to good. A careful estimate can be more informative than an experiment with a poorly described comparison. A story can contain several types, provided their scopes remain separate.
Read the evidence label with the original source and date. Ask whether the headline stays within the evidence, or quietly turns a plan into a result. Evidence type alone supplies no confidence percentage or QRank points.
Name the task, its input and required output quality. State the success rule, the resources counted and the baseline used. Missing information should remain visible.
A faster subroutine may still belong to a slower complete workflow. A resource estimate is different from a measured machine. Preparation, classical processing, queueing and repeated attempts matter when the claim concerns the whole task.
Keep physical qubits, logical qubits and runtime in their original units. If the tasks or measurement boundaries differ, explain both results side by side. A forced winner table would obscure the difference.
An assumption record connects a conclusion to its operating conditions, resource model and workflow boundary. When the reviewed source gives a value, preserve its unit and precision and identify the conclusion that depends on it.
Keep what was observed, what was assumed and what remains unreported separate. Unknown error rates are not zero errors. An editorial judgment is not a numerical probability without a defensible model.
If later evidence changes an assumption, describe which conclusion changes with it. Preserve the earlier version so readers can follow the reasoning. A useful record makes uncertainty inspectable.
A company release, a partner’s version and news coverage can all describe one underlying event. They may help verify wording or context without becoming separate experiments.
Group sources by the announcement or research they originate from. Identify what each source adds: an original observation, corroboration or interpretation. A different website does not by itself establish independent evidence. Versions of a paper belong to the same research family.
When sources conflict, preserve the earlier value, the corrected value and the reason. Secondary reporting can be useful; the point is to make its contribution clear rather than inflate a source-count badge.
A correction fixes an inaccurate statement. A clarification narrows ambiguity. New evidence can change an interpretation, and a status update can change an agreement or opportunity. Those changes should be named precisely; a spelling edit is different.
Keep the original publication date. Record what materially changed, why it changed and which source supports it. Retain the paper version originally cited and link any successor version rather than silently replacing it.
Present the accurate current conclusion prominently. A withdrawal needs a dated explanation and preserved provenance. Update affected summaries and cards so an old optimistic teaser does not survive a consequential correction.
Write the precise claim, task and output-quality target, source version, actual comparison and one important conclusion not established. Use the source-check worksheet for a worked reading process, the comparison worksheet for unlike hardware claims and the error-correction guide for encoded resources.
This source-locator check adds context to the existing record. Its original publication and claim-review dates remain unchanged. No independent scientific replication was performed.
How these sources are related
IBM wrote both accounts of its prototype. The August 19 announcement links to the information sheet on Zenodo. That repository hosts IBM’s document. The two links describe one demonstration.
Source relationship reviewed .
Evidence context
Company-reported cryogenic prototype
Independent replication
not assessed. This describes what was assessed, rather than evidence that replication failed.
Catherine Dundon; Shawn Hall; Matthew Hollister; Allie Lindler. IBM’s new modular architecture for cryogenic systems.
IBM Quantum Computing Blog. Source published 2026-08-19. Dated company report, 19 August 2026.
Why isn’t IBM using existing modular approaches for cryogenics? Final prototype paragraph. https://www.ibm.com/quantum/blog/modular-cryogenics#:~:text=two%20modular%20cryogenic%20cell%20prototypes
Source authority: IBM’s report that two prototype cryogenic cells were coupled and operated together.
Does not establish: Independent validation, a working inter-processor logical operation or completion of the roadmap.
Source relationship reviewed 2026-11-03: IBM wrote both accounts of its prototype. The August 19 announcement links to the information sheet on Zenodo. That repository hosts IBM’s document. The two links describe one demonstration.
QubitWire: Two cryogenic cells are an infrastructure demonstration. Record first published 2026-09-26. Cited revision claim-ibm-modular-cryogenic-prototypes-r1.
https://qubitwire.com/evidence/ibm-modular-cryogenic-prototypes?revision=claim-ibm-modular-cryogenic-prototypes-r1
Current record and corrections: https://qubitwire.com/evidence/ibm-modular-cryogenic-prototypes
Limitation: Cooling connected cells does not demonstrate fault-tolerant computation across chips. Deployment timing and future qubit capacities are plans, not measured computing outcomes.
Stefan Woerner; Robert Davis. QOBLIB: tracking progress in quantum optimization.
IBM Quantum. Source published 2026-08-12. Dated project update, 12 August 2026.
What is QOBLIB? First paragraph; subsequent project website discussion. https://www.ibm.com/quantum/blog/qoblib#:~:text=The%20Quantum%20Optimization%20Benchmarking%20Library
Source authority: The project’s stated ten problem classes, metrics, baselines and benchmark resources.
Does not establish: A universal optimization speedup or independent validation of results submitted to the library.
QubitWire: A shared benchmark library defines the comparison. Record first published 2026-09-26. Cited revision claim-qoblib-shared-benchmark-resource-r1.
https://qubitwire.com/evidence/qoblib-shared-benchmark-resource?revision=claim-qoblib-shared-benchmark-resource-r1
Current record and corrections: https://qubitwire.com/evidence/qoblib-shared-benchmark-resource
Limitation: A benchmark library is not evidence that quantum optimization has achieved practical advantage. This source check does not audit individual submissions or certify their reproducibility.
This source-locator check adds context to the existing record. Its original publication and claim-review dates remain unchanged. No independent scientific replication was performed.
How these sources are related
Four authors of the portfolio paper later summarized it on the AWS blog. Section 1 and reference 1 identify the original study. The hardware experiments in later sections belong to separate papers.
Source relationship reviewed .
Evidence context
Reported classical numerical benchmark
Independent replication
not assessed. This describes what was assessed, rather than evidence that replication failed.
Atithi Acharya et al.. Decomposition Pipeline for Large-Scale Portfolio Optimization with Applications to Near-Term Quantum Computing.
arXiv. Source published 2024-11-20. arXiv:2409.10301v2, 20 November 2024.
Abstract; Introduction, Our contribution; IV Numerical Results. https://arxiv.org/html/2409.10301v2#S4
Source authority: The tested decomposition method and its comparison with the same classical solver.
Does not establish: Quantum speedup, investment returns or identity with every result in the related journal version.
Source relationship reviewed 2026-11-03: Four authors of the portfolio paper later summarized it on the AWS blog. Section 1 and reference 1 identify the original study. The hardware experiments in later sections belong to separate papers.
QubitWire: Smaller portfolio subproblems improve a classical baseline. Record first published 2026-09-26. Cited revision claim-portfolio-decomposition-classical-baseline-r1.
https://qubitwire.com/evidence/portfolio-decomposition-classical-baseline?revision=claim-portfolio-decomposition-classical-baseline-r1
Current record and corrections: https://qubitwire.com/evidence/portfolio-decomposition-classical-baseline
Limitation: Results depend on the tested data and constraints. Potential compatibility with smaller quantum devices does not demonstrate quantum speedup or production portfolio returns.
Record first published 2026-10-08 · Revision claim-qbi-stage-c-selection-2026-10-07-r1.
Inspect the exact evidence and citation
This source-locator check adds context to the existing record. Its original publication and claim-review dates remain unchanged. No independent scientific replication was performed.
Evidence context
Programme status; future hardware evaluation
Independent replication
not assessed. This describes what was assessed, rather than evidence that replication failed.
DARPA. Four more teams enter QBI’s final stage.
DARPA. Source published 2026-10-07. Dated programme announcement, 7 October 2026.
Stage C description; selected-company list; paragraph explaining the programme is not a competition. https://www.darpa.mil/news/2026/four-more-teams-enter-quantum-benchmarking-initiative-final-stage#:~:text=The%20following%20companies
Source authority: DARPA’s Stage C selections and the stated scope of the evaluation.
Does not establish: A completed utility-scale machine, a technical ranking or investment performance.
QubitWire: DARPA is testing roadmaps, not declaring winners. Record first published 2026-10-08. Cited revision claim-qbi-stage-c-selection-2026-10-07-r1.
https://qubitwire.com/evidence/qbi-stage-c-selection-2026-10-07?revision=claim-qbi-stage-c-selection-2026-10-07-r1
Current record and corrections: https://qubitwire.com/evidence/qbi-stage-c-selection-2026-10-07
Limitation: Selection does not establish a successful useful application, completed fault-tolerant system or technical winner.
Timothy Proctor et al.. Benchmarking the computational power of quantum computers · v1.
Proctor and colleagues · arXiv. Source published 2026-09-10. arXiv:2609.12146v1, 10 September 2026.
Abstract, paragraphs 2–4; submission history for the cited v1. https://arxiv.org/abs/2609.12146v1#:~:text=Here%20we%20demonstrate%20direct
Source authority: The paper’s stated benchmark scope, named processors and separation of measurements from projections.
Does not establish: Independent reproduction, any individual uninspected table value or a universal platform winner.
QubitWire: QUOPS separates measured capability from future projections. Record first published 2026-10-08. Cited revision claim-quops-method-2609-12146-v1-r1.
https://qubitwire.com/evidence/quops-method-2609-12146-v1?revision=claim-quops-method-2609-12146-v1-r1
Current record and corrections: https://qubitwire.com/evidence/quops-method-2609-12146-v1
Limitation: No independently reproduced benchmark, universal platform winner or replacement QRank metric is established here.
The provider index dates the post 19 August and identifies general availability from 12 August. Direct retrieval timed out; the live product documentation was read.
Direct full text retrieval was blocked. The official PubMed record provides the author abstract and the 5 August publication date; no new numerical hardware claim is introduced here.
