Living Review Archive

Archive status: the first AMO frontier-map snapshot, amo-frontiers-2026-07-26, is frozen below. It records the evidence classification and literature cutoff used by the ten detailed frontier pages on 26 July 2026. The live pages may later change; this entry must not be silently rewritten to make the earlier assessment look current.

Scientific cutoff: version-of-record literature and explicitly labelled preprints available by 26 July 2026. A paper published after that date belongs to a later snapshot unless it corrects evidence that was already in scope.

Purpose and Canonical Boundary

A living review is useful only if readers can tell what changed. Replacing last year’s record values, platform rankings, or publication labels in place can make the current page cleaner while erasing the history needed to audit a claim. This archive preserves that history.

This page is the canonical home for:

annual AMO frontier maps;
snapshot identifiers and evidence cutoffs;
cross-program status summaries;
source-integrity manifests;
archive-level corrections and supersession records; and
the procedure for freezing the next annual map.

It is not the canonical home for atomic, molecular, optical, or quantum derivations. It does not preserve a second live copy of every frontier page. The detailed pages own current, topic-specific assessments, while their background links own enduring theory:

Frontiers and Open Problems owns the live chapter map and claim taxonomy.
AMO Bibliography and Reading Guide owns the volume-wide reading structure.
AMO Experiment Index owns the cross-platform experiment lookup.
Versioning and Review Policy owns global page status, review, and staleness policy.
Reference Changelog and Site Errata own broader correction records.

An archive entry may repeat a dated conclusion because preserving that conclusion is its purpose. It should not repeat a derivation, a full bibliography, or a current numerical table that already has a canonical home.

Snapshot Semantics

A snapshot is an as-of claim

The statement

amo-frontiers-2026-07-26 classified a platform claim as active

means that the cited evidence available at the declared cutoff supported that classification. It does not mean the classification remained correct for all later dates.

Each snapshot distinguishes three times:

Time field	Meaning
evidence cutoff	latest publication or public record admitted to the review
freeze date	date on which the page set, labels, and manifest were fixed
correction date	date on which an error in the frozen record was documented

A correction can be later than the freeze date without changing the evidence cutoff. A later scientific result normally creates a new snapshot rather than a correction.

Claim labels are preserved with their scope

The 2026 map uses the following labels:

Label	Archived meaning
established	directly demonstrated or standard within the claim’s stated scope
active	supported research program with central capability, scale, or interpretation still developing
conjectural	plausible extrapolation or model-dependent claim lacking decisive evidence
speculative	scientifically discussable proposal with a long or unvalidated evidence chain
controversial	live disagreement about evidence, interpretation, or sufficiency
not established	an explicit negative boundary; not a claim that the result is impossible

The noun governed by the label must be retained. “A gate primitive is established” and “fault-tolerant computation is established” are different claims even when they refer to the same apparatus.

Publication state is part of the snapshot

A preprint, accepted manuscript, version of record, correction, erratum, and reanalysis are different evidence objects. The snapshot preserves which one was available at the cutoff. Later peer review can strengthen a claim, weaken it, or leave its scientific content unchanged; it does not retroactively make the earlier source a version of record.

Negative findings are first-class records

The archive records absences that constrain interpretation:

no confirmed new force or dark-matter field;
no permanent EDM or new CP-odd nuclear moment;
no demonstrated general-purpose fault-tolerant AMO computer;
no universal quantum-simulation advantage;
no arbitrary chemical-reaction control; and
no model-independent attosecond “movie” of a many-electron process.

Deleting these boundaries from a later summary would create survivorship bias toward positive milestones.

Archive Entry Contract

Required metadata

Every annual entry should contain:

Field	Requirement
`snapshot_id`	stable, unique identifier containing scope and freeze date
`title`	human-readable map title
`evidence_cutoff`	inclusive date and any source-state qualification
`frozen_at`	freeze date or RFC 3339 timestamp
`routes`	complete set of frontier pages covered
`claim_vocabulary`	version or explicit definitions of evidence labels
`source_revision`	repository revision when available
`source_sha256`	exact-byte digest for each archived source
`renderer`	Astro, Starlight, KaTeX, and relevant integration versions
`verification`	build, rendered-math, route, and search-index result
`supersedes`	previous snapshot identifier, if any
`superseded_by`	next snapshot identifier, once frozen
`corrections`	append-only correction identifiers

A compact machine-readable representation can use:

snapshot_id: 'amo-frontiers-2026-07-26'
title: '2026 AMO Frontier Map'
evidence_cutoff: '2026-07-26'
frozen_at: '2026-07-26'
claim_vocabulary: 'amo-frontier-labels-1'
source_revision: null
source_hash_algorithm: 'sha256'
renderer:
  astro: '7.1.3'
  starlight: '0.41.4'
  katex: '0.17.0'
verification:
  production_build: 'pass'
  pages_built: 2139
  rendered_math_errors: 0
  broken_snapshot_routes: 0
supersedes: null
superseded_by: null
corrections: []

The 2026 entry records source hashes because a repository revision identifier was not available for this workspace freeze. Future entries should record both a revision and hashes.

Integrity is not scientific validation

For exact source bytes $B_i$ , the manifest records

h_i = \operatorname{SHA256}(B_i).

Matching $h_i$ shows that the bytes have not changed relative to the manifest. It does not show that the prose, mathematics, citation, or interpretation is correct. Scientific review, source criticism, and rendered verification remain separate checks.

Line-ending conversion, Unicode normalization, or a one-character edit changes the digest. A verifier must hash the exact archived bytes, not text copied from the browser.

A snapshot contains a map, not a hidden fork

The archive freezes:

the set of frontier routes;
the review cutoff;
each route’s broad established, active, and unestablished boundaries;
the year’s most interpretation-relevant changes;
negative findings;
source hashes and renderer metadata; and
later corrections.

It links to detailed pages for the full evidence and to canonical background for derivations. When a live page is superseded, its old full source can be moved under frontiers/living-review-archive/ with status: 'historical' and knowledge_status: 'historical'.

2026 AMO Frontier Map

Snapshot metadata

Field	Frozen value
snapshot identifier	`amo-frontiers-2026-07-26`
title	2026 AMO Frontier Map
evidence cutoff	26 July 2026
detailed frontier pages	10
claim vocabulary	`amo-frontier-labels-1`
source revision	not recorded; exact-byte SHA-256 manifest supplied
renderer baseline	Astro 7.1.3, Starlight 0.41.4, KaTeX 0.17.0
production verification	2,139 pages built; Pagefind indexed 2,139 HTML files; sitemap generated
previous snapshot	none
archive-level corrections at freeze	none

Frontier status map

Frontier	Established baseline at the cutoff	Active boundary
Precision AMO	coherent transduction, clocks, magnetometers, and atom interferometers	networks, long baselines, moving platforms, entanglement assistance, and new-physics interpretation
Optical clocks	lattice and ion clocks, frequency combs, and phase-stabilized fibre comparison	clock networks, transportable geodesy, entanglement, nuclear-clock systematics, and dark-field searches
Cold molecules	selected direct cooling and trapping, ultracold assembly, microwave shielding, molecular degeneracy, and pair entanglement	larger polyatomic cycles, defect-free arrays, dipolar phases, molecular qubits, and symmetry measurements
Rydberg arrays	thousand-atom storage arrays, hundred-atom simulators, parallel gates, logical encodings, and erasure-aware ingredients	repeated correction, integrated fault tolerance, calibrated optimization, and useful algorithmic scale
Ultracold-atom quantum simulation	Hubbard realizations, site-resolved microscopy, antiferromagnetic correlations, topological bands, and one-dimensional Abelian gauge mappings	low-temperature doping, interacting topology, higher-dimensional gauge theory, synthetic dimensions, and nonequilibrium validation
Quantum optics	antibunched and heralded photons, interference, squeezing, Gaussian processing, non-Gaussian preparation, and remote-memory entanglement	efficient indistinguishable sources, repeaters, integrated loss tolerance, optical GKP resources, and network advantage
Cavity and circuit QED	coherent single-emitter coupling, enhanced emission, dispersive control, collective radiance, ultrastrong coupling, and bosonic storage	multimode interfaces, delayed feedback, corrected bosonic memories, transduction, and modular links
Attosecond and ultrafast science	high-harmonic generation, attosecond pulses, streaking, RABBITT, pump–probe spectroscopy, diffraction, and strong-field ionization	quantitative charge migration, coupled electron–nuclear reconstruction, condensed-phase HHG, X-ray methods, and quantum-light driving
Molecular control	coherent transfer, alignment and orientation, selected pathway interference, feedback shaping, and prepared ultracold-state control	near-unit transfer, product-resolved reactions, single-molecule shaping, response-aware optimization, and nonadiabatic steering
Fundamental symmetries	weak parity violation, atomic parity violation, and calibrated reversal methods	EDMs, NSD-PV, Schiff and magnetic-quadrupole moments, radioactive molecules, isotope shifts, and new-boson searches

Exact-byte source manifest

The following SHA-256 values were computed from the ten UTF-8 source files in the workspace at freeze. They include the files’ original line endings.

Show source hashes

attosecond-and-ultrafast-frontiers.mdx
  dd82cebc71d52a968b2206155acd06c13cfafcd3dfe68ff3bb271d3154d17c29
cavity-and-circuit-qed-frontiers.mdx
  59c28ccfbeb0cd8ddc7a64f22149e8352fd8414e95212202b6713704ed8f83e9
cold-molecule-frontiers.mdx
  406e0317c609e17242b9b63f5e1eb132c3d9db95c2e2510ac5ebfe909b6e1e2b
fundamental-symmetry-frontiers.mdx
  6e8ba0141abb3b8851a455214aa5627395ffe91186686e544f7937654ab35873
molecular-control-frontiers.mdx
  d7db92b9493e5797dd921d491610c5d40c240f3dead14cd562703c71cae7d77a
optical-clock-frontiers.mdx
  c3cf2d1dd2cdeb55543cbfbfd68dabf66351916193532581231a5098a845264f
precision-amo-frontiers.mdx
  778832b37bb9fafd098727b0ff7ec70a09287596cc32a59f30430a01dfbbd0f2
quantum-optics-frontiers.mdx
  17da816a2f5dc120a766d84395abd90dd8fb1b8e03a11e78ba1a4b3696645b19
rydberg-array-frontiers.mdx
  e8be1424cac6a19fa606d0817f754a4c47c6c7ceb4a66e7abfb6c476eb68a15a
ultracold-atom-quantum-simulation.mdx
  e429db9a2cfba110bf34149d66326e9534ad4ce6504bc52d217764845d20a8f8

2026 Cross-Program Baseline

The 2026 review did not rank frontiers by publicity, funding, qubit count, frequency precision, or a single record. It adopted common evidence rules so that unlike platforms could be compared without pretending they solve the same problem.

Preserve the inference chain

Across the ten pages, claims were organized as

\text{raw record} \longrightarrow \text{calibrated observable} \longrightarrow \text{system response} \longrightarrow \text{scientific coefficient} \longrightarrow \text{model interpretation}.

Each arrow adds assumptions. A detected photon pattern is not yet a many-body phase; a molecular frequency is not yet an electron EDM; a transduced microwave tone is not yet a transferred arbitrary quantum state; and a nonlinear isotope shift is not yet a fifth force.

The archive retains the narrowest statement directly supported by the measurement and labels later interpretations separately.

Scale is a resource vector

The 2026 map rejected single-number platform rankings. Representative resource vectors were:

Program	Resources that had to be considered together
clocks and sensors	instability, systematic uncertainty, uptime, link performance, coverage, and deployability
neutral-atom arrays	loaded atoms, connected interactions, gate fidelity, loss, erasure conversion, decoding, and repeated cycles
molecular arrays	state purity, filling, shielding, coherence, addressability, and readout
quantum networks	source probability, coupling, conversion, transmission, detection, memory lifetime, and accepted-event definition
quantum simulation	preparation temperature, Hamiltonian calibration, observable access, finite-size control, and classical cross-check
ultrafast reconstruction	pulse characterization, response model, inversion uniqueness, gauge and basis dependence, and timing calibration
symmetry tests	coherence, reversal leverage, response coefficients, systematics, confidence construction, and operator complementarity

A larger value of one component can expose a deficit in another. More atoms increase neither useful computation nor simulation validity when loss, calibration, or readout is uncontrolled.

Platform primitive, integrated system, and scientific outcome are distinct

The review used a three-level evidence ladder:

primitive: a gate, interface, state-preparation step, pulse, transition, source, detector, or calibration;
integrated system: several primitives operating together with an end-to-end benchmark; and
scientific outcome: a validated answer to the claimed physics or information-processing question.

Examples preserved in the snapshot include:

a molecular clock transition that suppresses field sensitivity is a control primitive, not an EDM limit;
a kilometre transduction link carrying a coherent signal is an integrated interface milestone, not arbitrary quantum-state transfer;
fault-tolerant ingredients in a neutral-atom array are stronger than isolated gates, but do not yet establish sustained useful fault-tolerant computation;
a self-bound molecular droplet is a many-body observation, while supersolidity requires additional coherence and order evidence; and
an attosecond reconstruction can establish an observable-dependent delay without producing a unique classical trajectory of an electron.

Error bars do not absorb model inadequacy

The 2026 pages separated:

statistical uncertainty;
calibrated systematic uncertainty;
theory-response uncertainty;
finite-size or truncation error;
model discrepancy;
selection and trial effects; and
publication-state uncertainty.

Adding these as if they were independent Gaussian variances is not always valid. A missing Hamiltonian term, a nonidentifiable inverse problem, or an unrecognized source of King-plot curvature is not repaired by widening a statistical error bar without a model.

Reproducibility means repeating the relevant claim

The required replication depends on the claim:

Claim	High-value replication
record clock ratio	independent clocks, links, epochs, and ratio closure
logical neutral-atom operation	repeated cycles with loss, erasure, decoder, and feed-forward included
quantum-simulation phase	independent observable, system size, preparation path, and classical limit
remote entanglement	end-to-end heralding, memory survival, and multi-segment extension
chemical control	predicted product-resolved change under a prespecified control
EDM or new-force candidate	distinct apparatus with a complementary response coefficient
ultrafast movie	independent pulse characterization and an inversion that tests model alternatives

Repeating the same analysis on more data can reduce noise while leaving a shared model error untouched.

2026 Change Ledger

This ledger is selective. It records developments that changed an evidence boundary or clarified what a headline did not establish. Full references remain on the linked frontier pages.

Frontier	Change admitted by the cutoff	Archived interpretation
Precision AMO	Al $^+$ /Yb, Al $^+$ /Sr, and Yb/Sr ratios reached total uncertainties at or below $3.2\times10^{-18}$ ; the AION prototype demonstrated differential clock-transition interferometry under large synthetic laser noise	ratio reproducibility and long-baseline ingredients advanced; no new constant variation, dark-matter field, or gravitational wave was detected
Optical clocks	multispecies ratios improved; a solid-state thorium reference reported a temperature-insensitive region and cross-crystal reproducibility; two nuclear-clock feedback demonstrations appeared as preprints; a $2067\ {\rm km}$ optical link advanced dissemination	nuclear-clock operation and network engineering moved forward, but a complete nuclear-clock uncertainty budget and a global optical time network were not established
Cold molecules	microwave-dressed NaCs formed self-bound droplets and arrays; a ground-state NaRb gas formed a Bose–Einstein condensate of roughly 500 molecules and a tunable droplet	molecular degeneracy and droplet formation became multi-platform observations; crystalline order, supersolidity, and broad shielding universality remained active
Rydberg arrays	a neutral-atom fault-tolerance architecture entered the version of record with up to 448 atoms; two $^{171}$ Yb experiments joined erasure conversion, logical operation, and adaptive ingredients	many fault-tolerant components coexisted at meaningful scale; sustained useful general-purpose fault tolerance remained active
Ultracold-atom simulation	dimensional crossover was mapped with temperature as a control axis; bound-cluster tunnelling produced a massive spatial superposition; the 2025 low-temperature Hubbard result became the central cooling benchmark	state preparation and observable reach improved; broad quantum advantage and full higher-dimensional gauge dynamics were not established
Quantum optics	remote ion–ion entanglement through $10\ {\rm km}$ of fibre survived beyond the mean establishment time; integrated twin-field QKD expanded to 20 client chips; continuous-variable generation, gates, and readout were monolithically combined	repeater-relevant memory and integrated networking crossed important boundaries; a multi-segment repeater and general quantum internet did not
Cavity and circuit QED	a cavity-array microscope interfaced atoms across more than 40 free-space modes; a one-kilometre cryogenic photonic link preserved coherent signals; a superconducting Rice–Mele waveguide produced tunable directional edge states	parallel interfaces, long links, and directional coupling advanced; arbitrary quantum-state transfer and modular fault-tolerant networking remained unshown
Attosecond and ultrafast science	ion–photoelectron entanglement in dissociating $\mathrm{H_2^+}$ became a control parameter; liquid high harmonics gained trajectory-resolved timing; quantum statistics entered isolated-atom tunnelling; the shortest-pulse claim acquired a reconstruction dispute	new observables and control were established, while universal tunnelling time and model-independent trajectory claims remained unsupported
Molecular control	matter-wave chemistry showed phase doubling and nonclassical correlations; resonant atom–ion charge exchange showed interference across many partial waves; an open vibronic simulator gained a second engineered bosonic mode	coherence reached new reaction and simulation regimes; arbitrary product-selective chemistry and universally predictive short-range phases did not
Fundamental symmetries	the first direct deuteron EDM limit was published; YbF demonstrated a complete ultracold-beam spin interferometer; YbOH demonstrated engineered field-insensitive transitions; radioactive-molecule and new-boson analyses advanced	storage-ring and molecular control readiness improved; no permanent EDM, new CP-odd nuclear moment, molecular weak energy difference, or fifth force was observed

Developments retained from 2025

Several 2025 results remained the 2026 baseline because no later version-of-record result superseded them. Examples include calcium multitransition isotope-shift constraints, AcF and RaF spectroscopy, optical-GKP generation, molecular STIRAP benchmarks, bosonic-memory experiments, and low-temperature fermionic Hubbard preparation.

“What changed this year” therefore does not mean “only papers dated 2026 matter.” It records changes in the frontier assessment. A 2025 result can remain the controlling evidence in the 2026 map.

Corrections attached to cited papers were included

Where a source had a publisher correction or erratum by the cutoff, the corrected record governed the snapshot. Examples include the correction associated with the neutral-atom fault-tolerance article and the established erratum to the $^{199}$ Hg EDM paper. Consulting only the original article without its correction is not a faithful reconstruction of the 2026 review.

2026 Negative Findings and Open Claims

The snapshot’s most important restraint is often the sentence after a milestone.

Frontier	Not established at the 2026 cutoff
precision AMO	confirmed variation of a constant, dark-matter signal, new force, or long-baseline observatory detection
optical clocks	demonstrated superiority of a nuclear clock, full global optical time distribution, or new-physics discovery
cold molecules	universally transferable shielding, scalable defect-free molecular computation, molecular supersolidity, or symmetry violation
Rydberg arrays	sustained general-purpose fault-tolerant computation or practical optimization advantage under full resource accounting
ultracold-atom simulation	universal quantum-simulation advantage, complete finite-size removal, or generic higher-dimensional gauge-theory realization
quantum optics	practical multi-segment repeater, loss-tolerant general photonic processor, or universal quantum-network advantage
cavity and circuit QED	loss-tolerant remote processor interconnect, arbitrary-state kilometre transduction, or operational advantage from an ultrastrong-coupling ground state
attosecond and ultrafast science	universal tunnelling time, unique many-electron trajectory reconstruction, or unlimited sub-attosecond temporal resolution
molecular control	routine arbitrary reaction control, universally reliable optimal-control transfer, or complete predictive nonadiabatic dynamics
fundamental symmetries	permanent EDM, Schiff moment, nuclear magnetic quadrupole moment, molecular weak enantiomeric energy difference, or fifth force

Active does not mean equally near completion

An active program can be:

improving a mature null experiment;
assembling the first complete instrument;
validating one missing primitive;
resolving a theory uncertainty;
scaling an integrated system;
testing a disputed interpretation; or
converting a proposal into a measured observable.

The label records a live, evidence-bearing program. It does not estimate a calendar date for success.

Conjectural projections remain conditional

The 2026 pages retained conjectural claims when they were useful and clearly labelled, including:

future energy-scale reach inferred from a low-energy null result;
fault-tolerant or optimization advantage extrapolated from primitives;
nuclear-clock superiority before complete systematic evaluation;
network advantage projected from component efficiencies;
chemical selectivity projected from an optimized model; and
new-particle interpretations that set other effective operators to zero.

The archive preserves the assumptions that make each projection meaningful. It does not promote a projection because a later press summary used declarative language.

Disputed claims require an issue, not a vague warning

For a controversial classification, the archive should state what is disputed:

reconstruction uniqueness;
statistical coverage;
Hamiltonian fidelity;
resource accounting;
publication status;
theory response;
finite-size extrapolation; or
whether the measured observable supports the advertised noun.

“More work is needed” is too vague to reconstruct the disagreement.

Corrections and Supersession

Correction and update are different operations

A correction repairs the frozen record because the record was wrong at its own cutoff. Examples are a misquoted number, omitted erratum, broken route, incorrect publication label, sign error, or claim that exceeded the available evidence.

An update incorporates evidence that became available after the cutoff. It belongs to the next snapshot.

The distinction can be written as

\begin{aligned} S_{2026} &\xrightarrow{\text{error about 2026 evidence}} S_{2026}+E_1,\\ S_{2026} &\xrightarrow{\text{new 2027 evidence}} S_{2027}. \end{aligned}

Here $E_1$ is an append-only correction record. It does not erase the original snapshot identifier.

Required correction record

An archive correction should state:

Field	Meaning
correction identifier	stable identifier such as `amo-2026-E1`
recorded date	when the correction was published
affected snapshot and route	exact historical scope
issue	what the snapshot said or omitted
corrected statement	replacement wording or data
evidence	source that demonstrates the error
scientific impact	none, local, status-changing, or cross-program
hash impact	hashes or archived files affected

If the correction changes a scientific status label, the old label and new label must both remain visible.

Supersession preserves both directions

When the 2027 map is frozen:

the 2026 entry gains superseded_by: amo-frontiers-2027-...;
the 2027 entry gains supersedes: amo-frontiers-2026-07-26;
live frontier pages point to the newest assessment;
historical citations can still target the 2026 anchor; and
no 2026 negative finding is deleted merely because a later experiment succeeds.

An individual frontier page may be superseded before the annual map if a major correction or discovery makes its live assessment unsafe. Preserve the old source under the archive route and record the exceptional freeze.

Current correction log

amo-frontiers-2026-07-26: no archive-level correction was recorded at freeze.

This statement is itself dated 26 July 2026. Future corrections should be appended immediately below it with stable identifiers; they should not alter the sentence to imply that no correction was ever needed.

Annual Rollover Workflow

1. Declare scope before searching

Fix:

the frontier routes under review;
the inclusive evidence cutoff;
accepted source states;
claim-label definitions;
whether cross-volume developments are in scope; and
the date by which unresolved source checks defer to the next cycle.

Changing the scope after seeing an exciting result creates an inconsistent map. An exceptional late paper can be included only if the cutoff changes for the whole snapshot and the affected searches are rerun.

2. Audit the previous map

For every archived claim, ask:

Did a result become established, remain active, or weaken?
Was a projection tested?
Did a preprint enter the version of record?
Was a correction, retraction, or expression of concern issued?
Did theory alter the interpretation without changing the data?
Did a better canonical home appear?
Is a negative finding still true?

The audit starts from the old map rather than from a fresh list of recent papers. That preserves continuity and makes status transitions visible.

3. Search by claim and by platform

A reliable search uses two directions:

claim-first: search for the observable, capability, or disputed interpretation across platforms;
platform-first: search each major apparatus or method for results that may alter a neighboring claim.

For example, a new nuclear-structure calculation can change a molecular EDM interpretation even though the experiment did not publish new data. A network component can alter a quantum-repeater budget without itself being a repeater demonstration.

Prefer version-of-record papers, publisher corrections, standards, and authoritative reviews. Preprints can enter when they materially define the frontier, but their provisional state must be visible.

4. Update the evidence ledger

Each candidate change should record:

Item	Question
measured object	What was directly observed or fitted?
comparison	Relative to which baseline, model, or previous result?
uncertainty	Statistical, systematic, theory, and model components?
scope	Species, state, geometry, task, distance, duration, or parameter range?
publication state	Preprint, accepted paper, version of record, correction, or reanalysis?
claim effect	Which archived status sentence changes, if any?
canonical route	Where does the full explanation belong?

A paper can be important yet leave the map unchanged. Such a paper belongs in the detailed bibliography, not necessarily in the annual change ledger.

5. Review cross-program consistency

Before freezing, compare repeated nouns across pages:

“coherence” must identify the degree of freedom and observable;
“fidelity” must state the operation, conditioning, and loss treatment;
“quantum advantage” must name the task and resource accounting;
“network” must distinguish components, links, memories, and end-to-end operation;
“control” must identify the target outcome and validation;
“discovery” must distinguish measured anomaly from model interpretation; and
“record” must state the metric and comparison class.

Cross-page inconsistency is an archive-level problem even when each sentence looks plausible in isolation.

6. Freeze source and provenance

After scientific review:

stop substantive edits to the snapshot set;
record route and frontmatter metadata;
record repository revision when available;
compute exact-byte SHA-256 hashes;
record dependency versions from the lockfile;
create the human-readable map;
create machine-readable metadata when supported; and
assign supersedes and superseded_by relations.

If a file changes after hashing, recompute the manifest and rerun every verification affected by the edit. A manifest assembled from different source moments is not one snapshot.

7. Run acceptance checks

The minimum release gate is:

Check	Acceptance condition
content schema	all frontmatter validates
MDX	every covered page compiles
KaTeX	no error markup or red fallback math
internal routes	all snapshot and canonical links resolve
figures	referenced assets exist and have useful alternatives
production build	completes without a new relevant warning
rendered structure	expected title, sections, tables, and disclosures appear
sitemap	archive and live routes are included
search	snapshot identifier and distinctive claims are retrievable
manifest	recomputed source hashes match

Build success alone does not verify scientific accuracy. Scientific review alone does not verify that the archived object is renderable or retrievable.

8. Publish the map before resuming live edits

Append the new annual section, link it from the live frontier index, and record the old–new relation. Only then resume ordinary updates to the live pages.

Future annual sections should follow:

2026 AMO Frontier Map
2027 AMO Frontier Map
2028 AMO Frontier Map
...

When an annual section becomes too large for one page, preserve this route as the archive index and move each frozen map to a dated child route. Do not reuse a child route for a later year.

Mid-cycle exceptional freezes

A mid-cycle snapshot is warranted when:

a discovery claim materially changes several pages;
a major result is retracted or corrected;
a convention or response calculation changes cross-program inference;
a sidebar or canonical-home reorganization would make the old map unrecoverable; or
an archive integrity failure affects the frozen source.

Routine records and incremental improvements wait for the annual cycle. This keeps snapshot identifiers meaningful.

Integrity, Citation, and Reuse

Verify bytes and metadata separately

A complete verification has at least four independent questions:

Integrity: do the source bytes match their hashes?
Provenance: where, when, and by which workflow was the snapshot made?
Renderability: does the declared environment build the pages?
Scientific traceability: do the claims point to the right evidence and canonical explanation?

W3C PROV supplies a general vocabulary for entities, activities, and agents. The archive does not require a full PROV graph, but its metadata should be convertible to that model: a source page is an entity, review and build are activities, and the snapshot is a derived entity.

Hashes detect change, not authorship

SHA-256 is used here as a change detector. A bare digest does not establish:

who created the source;
whether the manifest is authentic;
whether the source was reviewed;
whether dependencies are trustworthy; or
whether the scientific claims are correct.

A repository revision, signed release, trusted publication record, or independent mirror can strengthen authenticity. Those mechanisms are additive rather than substitutes for claim-level review.

Cite the object that supports the sentence

For current scientific content, cite the live frontier page and its last_reviewed date.

For a historical statement about the review itself, cite the archive entry:

Living Review Archive, snapshot amo-frontiers-2026-07-26, evidence cutoff 26 July 2026, section “2026 AMO Frontier Map.”

For a primary scientific result, cite the original paper. The archive is evidence of how the result was classified, not a replacement for the paper.

Stable anchors matter

Annual headings and correction identifiers should remain stable. If the archive is split into child pages, preserve redirects or explicit migration links. A citation to a historical section should not silently land on the newest map.

If persistent external identifiers are added later, version metadata and relations such as IsNewVersionOf and IsPreviousVersionOf should accompany them. DataCite’s version and related-identifier fields provide a suitable general model.

FAIR does not mean frozen forever

The FAIR principles ask that digital research objects be findable, accessible, interoperable, and reusable. For this archive:

findable: stable route, snapshot ID, title, date, and search indexing;
accessible: ordinary web page plus source in the maintained project;
interoperable: explicit dates, routes, status vocabulary, hashes, and machine-readable metadata;
reusable: clear scope, provenance, correction policy, and links to primary evidence.

FAIR describes stewardship qualities. It does not require pretending that an old scientific assessment remains current.

Reuse must preserve the cutoff

A figure, table, or status sentence reused from a snapshot must retain:

snapshot identifier;
evidence cutoff;
claim scope;
source links;
correction state; and
any publication-state caveat.

Removing the date from a historical platform comparison turns an auditable snapshot into an unattributed current claim.

Common Failure Modes

Silently replacing the annual map

Changing “2026” values to “2027” values under the same anchor destroys the earlier record. Append a new map and link the versions.

Treating the archive as the current reference

The archive answers “what was the assessment at the cutoff?” The live frontier page answers “what is the current maintained assessment?”

Archiving only positive records

Milestones without negative findings systematically inflate maturity. Preserve what was not demonstrated and which projections remained conditional.

Copying canonical derivations into snapshots

This creates several stale homes for the same mathematics. Archive the dated conclusion and link to the enduring derivation.

Hashing rendered HTML but not source

Rendered HTML changes with templates, dependencies, and asset pipelines. Source hashes and renderer metadata answer different questions; preserve both when a full render archive is made.

Recomputing a hash after an undocumented edit

A matching new hash only proves consistency with the edited file. The edit still needs a correction or a new snapshot, depending on its cause.

Calling a late paper a correction

Evidence published after the cutoff normally belongs to the next snapshot. It corrects the old map only when it demonstrates that the map misstated evidence already available within scope.

Losing publication state

Replacing “preprint” with “paper” after publication makes the old assessment anachronistic. The next map can record the status transition.

Ranking unlike quantities

Qubit count, clock uncertainty, EDM magnitude, entanglement distance, pulse duration, and molecular temperature do not share an ordering relation. Each requires its own task and resource context.

Treating no correction as proof of correctness

An empty correction log means no archive-level correction has been recorded. It is not a guarantee that no error exists.

Exercises

Exercise 1: Correction or next snapshot?

Classify each event for amo-frontiers-2026-07-26:

A cited 2025 paper had an erratum published in May 2026, but the snapshot missed it.
A new experiment published in September 2026 improves a record.
A route in the 2026 map was misspelled.
A 2027 review reinterprets unchanged 2025 data with a new response calculation.

Solution

Correction. The erratum existed before the cutoff, so the frozen map failed to represent in-scope evidence.
Next snapshot. The result did not exist at the cutoff.
Correction. The historical route record is objectively broken. Record the old and fixed route and update affected hashes.
Next snapshot in the ordinary case. The new theory is post-cutoff evidence, even though the experimental data are older. If the new review merely exposes an arithmetic or citation error that was already demonstrable in 2026, that narrower issue can also receive a correction.

Exercise 2: What does a hash mismatch prove?

An archived file is recovered with Unix line endings, while the manifest was computed from Windows line endings. The prose is visually identical, but the SHA-256 value differs. What can be concluded?

Solution

The recovered bytes do not match the exact object that was hashed. The mismatch does not show that the scientific text changed, because line-ending conversion is sufficient to alter the digest.

Record the normalization history if it is known. Do not replace the old hash silently. A normalized-text digest can be added as a separate field in a future format, but it must not be confused with the exact-byte digest.

Exercise 3: Preserve the noun governed by a status label

In 2026, an array demonstrated logical gates and repeated syndrome information. A summary says “fault-tolerant neutral-atom computing was established.” Diagnose the archive problem and write a narrower statement.

Solution

The summary promotes evidence for several fault-tolerant ingredients into evidence for an integrated, sustained computing capability. The governed noun changed.

A defensible archive sentence is:

Several fault-tolerant neutral-atom primitives and integrated logical operations were established; sustained useful general-purpose fault-tolerant computation remained active.

The narrower statement preserves both the milestone and the missing system boundary.

Exercise 4: Which object should be cited?

You are writing:

“The current leading electron-EDM limit is …”
“The July 2026 frontier review classified YbOH clock transitions as a control milestone rather than an EDM result.”
“Takahashi et al. measured electric- and magnetic-sensitivity suppression in YbOH.”

Which source type belongs with each sentence?

Solution

Cite the current live Fundamental Symmetry Frontiers page and the primary limit paper. A historical snapshot may be stale.
Cite snapshot amo-frontiers-2026-07-26 on this page, because the sentence is about the review’s dated classification.
Cite the Takahashi et al. primary paper. The archive can document how the result was interpreted, but it is not the experimental source.

Exercise 5: Design a minimal annual freeze

You have ten reviewed pages, but no Git commit identifier. What is the minimum defensible freeze, and what limitation must be stated?

Solution

Record:

a unique snapshot ID;
evidence cutoff and freeze date;
complete route list;
claim-label definitions;
exact-byte hash for every source;
renderer and dependency versions;
build and rendered-verification results;
annual status and negative-finding summaries; and
correction and supersession fields.

State that no repository revision or signed release identifies the source state. The hash manifest can detect changes relative to itself, but without an independently trusted copy it provides weaker provenance and authenticity than a revision-backed release.

Connections

Live frontier assessments

Governance and reproducibility

Versioning and Review Policy defines review triggers, promotion, staleness, and changelog behavior.
Page Status Labels defines site-wide editorial and knowledge states.
Learn Changelog records reader-facing structural and convention milestones.
Reference Changelog and Site Errata record cross-site corrections.
Computational Reproducibility Benchmarks develops environment, artifact, convergence, and evidence practices for computational AMO work.
AMO Bibliography and Reading Guide and the AMO Experiment Index provide canonical source and platform discovery.

References

M. D. Wilkinson et al., “The FAIR Guiding Principles for Scientific Data Management and Stewardship,” Scientific Data 3, 160018 (2016), doi:10.1038/sdata.2016.18 — findability, accessibility, interoperability, and reuse of digital research objects.
T. Lebo et al., PROV-O: The PROV Ontology, W3C Recommendation (30 April 2013) — interoperable entities, activities, agents, and derivation relations.
National Institute of Standards and Technology, FIPS PUB 180-4: Secure Hash Standard (2015) — specification of SHA-2 message-digest algorithms used for change detection.
DataCite, Metadata Schema 4.6 (2025) — version, date, description, identifier, and related-resource metadata.
National Academies of Sciences, Engineering, and Medicine, Reproducibility and Replicability in Science, National Academies Press (2019) — distinctions among computational reproducibility, experimental replication, uncertainty, and evidential confidence.
Versioning and Review Policy — local canonical policy for page review, staleness, promotion, and changelog triggers.
Frontiers and Open Problems — live AMO claim taxonomy, reading protocol, and route map from which the 2026 snapshot was frozen.