Cosmological Tensions

Speculative If any one of these tensions is a real disagreement rather than an accounting error, then the six-parameter model that has fitted everything from the microwave background to the clustering of fourteen million galaxies is wrong in a specific, locatable place — the expansion history — and cosmology gets its first empirical handle on physics beyond that model since dark energy itself. The condition comes first because the audit is not finished, and section 7 says what would finish it.

Established The magnitude is a percentage, and it has been stable for a decade: the local and early-universe expansion rates differ by about 8 to 9%. The distance-ladder value is 73.49 ± 0.93 km s⁻¹ Mpc⁻¹ on the 2025 JWST-era calibration; the microwave-background inference is 67.4 ± 0.5. Closing that gap inside the standard model requires the sound horizon at recombination to shrink by roughly 5%, which means changing the expansion rate in the few hundred thousand years before the microwave background was emitted — a period no other observation constrains directly. Frontier The dark-energy claim is smaller and newer: a 3.1 standard-deviation preference for an equation of state that crosses w = −1, which if it holds means the accelerating component is not a constant and the universe’s long-term fate is a different question than it was in 2024.

Frontier Who gets it is unusually clear, and it is nobody in particular. There is no product at the end of this. The gain accrues to a field of perhaps a few thousand people, to the four or five survey collaborations whose instruments would be vindicated, and — genuinely, if diffusely — to the public record of what is known about the universe. Speculative The cost side is concrete where the benefit is not: the instruments that would settle these questions are national-scale facilities with decade lead times, and they are being built anyway for other reasons.

Handwave The condition, stated plainly: the upside is real only if the tensions survive their own audit, and the audit is going badly for the simple version of the story. A January 2026 reanalysis of 88 measurements finds the split is not early-universe against late-universe but distance-ladder against everything else; the second JWST programme reports a local expansion rate consistent with the microwave background and says no new physics is needed; and the dark-energy preference weakens to between 1.1 and 2.3 standard deviations when late-time data are used alone. Section 5 is about the fact that the binding constraint here is calibration, not physics.

The standard cosmological model has six free parameters. Fix them against the microwave background and it predicts, without further adjustment, the clustering of galaxies, the abundance of light elements, the growth of structure and the distance to a supernova at any redshift. A cosmological tension is the claim that one of those predictions has been checked and failed — that two independent determinations of the same quantity disagree by more than their stated uncertainties allow. That is a claim about error bars at least as much as it is a claim about the universe, which is why this subject belongs in the pre-paradigmatic category rather than in observational cosmology.

Established Four tensions are in play, and they are in very different states. The Hubble tension is a roughly 8 to 9% disagreement about the present expansion rate. The S₈ tension is a few-percent disagreement about the amplitude of matter clustering. The dark-energy question is whether the equation of state w evolves with redshift. And a fourth, newer anomaly is that fitting the neutrino mass sum to cosmological data drives it toward or below zero, against a laboratory lower bound from oscillation experiments. Frontier Only the first has ever been described as five sigma, and the description is doing more work than the number.

Established Two of the four quantities are not measured in the sense a reader would assume. The microwave-background value of the expansion rate is not an observation of how fast anything is receding; it is the number the standard model requires in order to reproduce the observed angular scale of the acoustic peaks. Speculative Change the physics before recombination and that inference changes without a single photon being remeasured. A tension between a direct measurement and a model inference is a different object from a tension between two measurements, and almost every summary of this subject conflates them.

Frontier The framing this page tests is that the standard model may be breaking, and it survives in a weaker form than the field’s own headlines suggest. Something is wrong: the numbers do not reconcile. Speculative But the most economical account of where the wrongness lives has moved, over 2025 and 2026, from “new physics before recombination” toward “the local calibration chain, or the way significance is being computed, or both”. Handwave Whether that is scientific caution or a field talking itself out of its own discovery is the argument this page is about, and no instrument settles it.

Established Begin with the anchor this brief was commissioned against, because it needs correcting in three places. The commission gives 73.0 ± 1.0 from the distance ladder against 67.4 ± 0.5 from the microwave background, about five standard deviations between two audited measurements, and 2.8 to 4.2 standard deviations from DESI DR2 for evolving dark energy. Established The arithmetic is right and the framing is four years old. The published ladder value it refers to is 73.04 ± 1.04 km s⁻¹ Mpc⁻¹, and against 67.4 ± 0.5 that is 4.9 standard deviations, which rounds to the five the commission states.

Frontier The first correction: there are not two audited local measurements, there are at least three, and they do not agree with each other. The Chicago–Carnegie Hubble Program, using JWST and HST on the same kind of galaxies, reports 70.39 ± 1.22 (stat) ± 1.33 (sys) ± 0.70 km s⁻¹ Mpc⁻¹ from the tip of the red giant branch, 68.81 ± 1.79 ± 1.32 from JWST-only red giants, and 67.80 ± 2.17 ± 1.64 from J-region asymptotic giant branch stars. Established Its authors state that their results are consistent with the standard model without the need for new physics, while also reporting that their distances agree with the Cepheid distances of the competing programme at just over the 1% level. Frontier Two teams, one telescope, overlapping galaxies, distances that agree to about a percent, and expansion rates that differ by about two standard deviations. Speculative That is the most important unresolved fact in this brief and it is not a disagreement about cosmology.

Established The second correction: the early-versus-late framing itself is now contested, by an audit rather than by a model. A January 2026 analysis assembled 88 sound-horizon-free determinations of the expansion rate and sorted them into distance-ladder measurements and everything else. The 30 distance-ladder measurements give 72.73 ± 0.39 with reduced chi-squared 0.72; the 58 ladder-independent, sound-horizon-free measurements give 69.37 ± 0.34 with reduced chi-squared 0.95. Frontier That is a 6.5 standard deviation discrepancy, larger than the canonical one, falling to a minimum of 3.9 standard deviations once correlations between the measurements are accounted for. Speculative The authors’ conclusion is that the tension is better characterised as distance ladder against all other methodologies than as early against late — which, if right, moves the suspect from the physics of recombination to the calibration of the ladder.

Frontier The same audit contains a result that should worry anyone quoting a single number. Within the non-ladder measurements, those that assume the standard model recover 67.61 ± 0.96, those independent of the shape of the expansion history recover 71.03 ± 0.69, and microwave-background analyses that avoid the sound horizon recover 69.07 ± 0.44. Established Imposing the standard model on local kinematic data lowers the inferred expansion rate by about 3.4 km s⁻¹ Mpc⁻¹, a 2.9 standard deviation internal tension that the authors themselves flag as sensitive to how they sorted the measurements. Handwave A field in which the answer depends measurably on which model you assumed while measuring is not yet in a position to announce new physics.

Established The independent methods, taken one at a time, do not referee. Strong lensing time delays, unblinded in May 2025 across eight lensed quasars with JWST, Keck and VLT kinematics, give 71.6 ^+3.9_−3.3 km s⁻¹ Mpc⁻¹, a 4.6% measurement sitting between the two camps and about 1.3 standard deviations from the microwave background. Established A distance-ladder-free expanding-photosphere analysis of ten type II supernovae between redshift 0.01 and 0.04 gives 74.9 ± 1.9, on the high side. Frontier Two gravitational-wave analyses published in 2026 disagree with each other by more than either disagrees with anything else: 142 events combined with weak lensing and galaxy clustering give 67.94 ^+4.40_−4.34 at 6.4% precision, while 17 dark sirens from the first part of the fourth observing run give 78.2 ^+12.0_−11.0. Speculative Every one of these is consistent with both camps, which is the honest summary of the independent-method programme in 2026. Multi-Messenger Astronomy owns why the siren route is so slow and this page does not restate it.

Established The dark-energy result is the newest and the best-measured, and the commission’s figure needs its attribution restored. DESI DR2 measured baryon acoustic oscillations in more than 14 million galaxies and quasars, reaching 0.24% statistical precision on the galaxy and quasar scale and 0.65% on the Lyman-alpha forest. Established Combined with the microwave background it prefers an evolving equation of state over a cosmological constant at 3.1 standard deviations; the 2.8 to 4.2 range the commission quotes is what happens when a supernova sample is added, and it runs from 2.8 with Pantheon+ through 3.8 with Union3 to 4.2 with the five-year Dark Energy Survey sample. Frontier The range is a measure of how much the answer depends on which supernovae you trust, not a measure of the strength of the evidence. The preferred solution sits in the quadrant with w₀ > −1 and w_a < 0.

Frontier The same data set carries an anomaly the collaboration reported and the coverage mostly did not. Fitting the sum of neutrino masses within the standard model produces a 3 standard deviation tension with the laboratory lower bound from oscillation experiments, pushing the cosmological preference toward or below zero. Established That tension disappears when the dark-energy equation of state is allowed to evolve. Speculative Two anomalies in one data set that cancel under one extension is either a strong hint or a warning that the extension has enough freedom to absorb whatever the data do, and the literature has not decided which.

Frontier The dark-energy preference has weakened under reanalysis, in two distinct ways. A 2026 study combining DESI DR2 with Pantheon+, the recalibrated DES-Dovekie sample and Union3 finds the statistical significance modest, at 1.1 to 2.3 standard deviations, with no model showing a robust preference for dynamical dark energy using late-time datasets alone. Frontier Separately, a 2025 critique argues that the combinations driving the DESI result are internally tense between the microwave background, the oscillation scale and the supernovae, and that no individual dataset independently detects cosmic acceleration at a significant level once the equation of state is allowed to evolve. Speculative The preference is real in the combined fit and fragile in every decomposition of it, which is the signature of a result driven by a calibration offset between datasets rather than by a feature in any one of them.

Established An independent survey reported in January 2026 and did not resolve it. The Dark Energy Survey’s six-year analysis combined all four of its probes — oscillation scale, type Ia supernovae, galaxy clusters and weak lensing — for the first time, with constraints more than twice as strong as its previous analyses. Frontier The data fit the standard model, and also fit an evolving dark-energy model but not better. Speculative A survey designed for exactly this question, reporting its full dataset, declined to confirm the headline of the year before.

Established The clustering-amplitude tension has not merely weakened; it has inverted, and this is the cleanest lesson in the brief. A 2026 review sets a unified baseline from Planck, ACT DR6 and SPT-3G of S₈ = 0.836 ^+0.012_−0.013, against Planck 2018 alone at 0.832 ± 0.013. Established Measured against that baseline, the KiDS Legacy weak-lensing analysis is statistically consistent at below one standard deviation, while Dark Energy Survey Year 6 sits at 2.4 to 2.7 standard deviations. Frontier The survey that for years supplied the S₈ tension now agrees with the microwave background, and the survey that used to agree now does not. Speculative The review’s own reading is that the heterogeneous landscape points to survey-specific systematics, without excluding new physics. A tension that swaps which instrument reports it, while keeping roughly the same size, is evidence about instruments.

Frontier The open questions here are not the ones the popular framing suggests, and five of them carry the subject.

Frontier First: is the local distance ladder wrong? The case that it is not is that multiple independent calibrators — Cepheids, red-giant tips, asymptotic-giant-branch stars, Miras, surface-brightness fluctuations — return consistent high values, and a single systematic would have to corrupt all of them coherently. Speculative The case that it might be is that the second JWST programme, using the same telescope and overlapping galaxies, returns values between 67.8 and 70.4 and says no new physics is needed. Handwave Both cases are made by people with a decade invested in the answer, and the disagreement has not narrowed in four years of new data, which is itself a datum about how it will be resolved.

Frontier Second: is the dark-energy signal a feature of the universe or of supernova photometry? The 1.4 standard deviation spread between the Pantheon+, Union3 and Dark Energy Survey combinations comes from samples that observe overlapping supernovae with different photometric systems. Speculative Cross-calibrating those systems is the single largest tractable systematic in the field, and the appearance of a recalibrated sample in the 2026 literature, with a weaker preference, is the first real test of that hypothesis. Handwave If the answer is photometry, then a result reported as the biggest cosmological news in a decade will have been a zero-point.

Frontier Third: what would a solution have to do? Any proposal that raises the local expansion rate by shrinking the sound horizon must do so without spoiling the acoustic peak structure, the light-element abundances, or the clustering amplitude. Speculative Early dark energy is the most studied candidate and it pays for a higher expansion rate with a higher clustering amplitude, which was tolerable when the clustering tension ran the other way and is less so now that the microwave-background baseline sits at 0.836. Handwave The solution space is narrow, heavily populated, and almost entirely unconstrained by any observation that was not already used to define the problem.

Speculative Fourth: is the significance being computed correctly? Every number in section 3 is a tension quoted for a comparison chosen after the data were seen, from a literature that generates hundreds of model extensions a year and reports the ones that fit. Handwave No standard convention exists for the trials factor across this field, and the January 2026 audit’s own headline number moves from 6.5 to 3.9 standard deviations depending on a correlation treatment its authors describe as sensitive to categorisation. A discipline whose central claim is a significance level and which has no agreed convention for computing one has a methodological problem prior to its physical one.

Speculative Fifth: what happens if nothing breaks? The modal outcome, on the evidence in section 3, is that the clustering tension dissolves into survey systematics, the dark-energy preference settles near two standard deviations and stops moving, and the expansion-rate disagreement persists at three to five standard deviations without either side conceding. Handwave That is not a resolution and it is not a crisis; it is a field with an unfalsified anomaly and no mechanism for retiring one, which is the condition this corpus calls pre-paradigmatic.

Frontier The bottlenecks in cosmological tensions are calibration bottlenecks, and naming them as such is the most useful thing this section can do.

Established Photometric cross-calibration between supernova surveys is the binding constraint on the dark-energy claim. The 2.8-to-4.2 spread is generated by samples that share objects and differ in zero-points, filter response functions and selection. Speculative Until one calibration is applied to all of them by a team with no stake in the outcome, the significance of evolving dark energy is not a measurement of the universe.

Frontier The absolute calibration of the distance ladder rests on a handful of geometric anchors. Parallaxes, one water-maser galaxy and detached eclipsing binaries in the Magellanic Clouds carry the zero-point that every rung above them inherits. Speculative The anchors are few enough that an error in one is not averaged away, and the 2026 Gaia data release is the next event that could move them. A three-rung ladder with three anchors is not a redundant system.

Speculative The sirens are supply-limited, not precision-limited. One neutron-star merger with a confirmed electromagnetic counterpart has been found in nine years, and that event’s expansion rate carries an error bar wide enough to be consistent with everything. Handwave Multi-Messenger Astronomy shows that the constraint is sky area, cadence and spectrograph queues rather than detector sensitivity, which means the siren route cannot be accelerated by building a better interferometer alone. The 2026 standard-siren review puts the requirement for a 1 to 2% expansion rate at roughly 100 to 1,000 counterpart events, against a delivered total of one.

Frontier The microwave background has no successor satellite. The definitive full-sky measurement is now more than a decade old and the improvements since have come from ground-based experiments covering partial sky at higher resolution. Speculative A tension whose early-universe leg has not been remeasured by an independent full-sky instrument since 2018 is a tension with an unaudited half, and no such mission is funded to fly. Mega-Telescopes carries the aperture economics that makes ground-based substitution attractive and incomplete.

Speculative The least technical bottleneck is analysis freedom. Blinding is now routine in the best analyses — the 2025 time-delay result unblinded on a stated date, the weak-lensing surveys blind their shear calibrations — and it is not universal, and there is no convention covering the choice of which datasets to combine. Handwave That choice is where the 2.8-to-4.2 range lives, and it is currently made by authors after they can see what each combination gives.

Frontier This subject is unusual in that its dependencies have dates, because the instruments exist and the data releases are scheduled.

Established Four releases land within about a year of this brief. Euclid’s first data release covers roughly 1,900 square degrees plus the deep fields in October 2026; the Rubin Observatory releases its science-grade commissioning data as a second data preview; the DESI collaboration publishes a second wave of DR2 analysis papers testing gravity against combined probes; and Gaia Data Release 4, containing 66 months of observations for 2.8 billion sources, arrives in December 2026. Frontier Gaia is the one that matters most to the anchor problem, because parallax zero-points propagate directly into the ladder.

Frontier The dependency that has no date is a recalibrated supernova compilation. The appearance of a cross-calibrated sample in the 2026 dark-energy literature is a first step; a single photometric system applied across all the major samples is not scheduled by anyone. Speculative It is the cheapest decisive work in this brief and it is nobody’s programme, because it is service work that can only reduce someone’s published significance.

Speculative Two dependencies run outside astronomy. The neutrino-mass anomaly is bounded from below by laboratory oscillation experiments and from above by direct kinematic limits, so a cosmological preference for a negative effective mass is adjudicated partly in particle physics rather than in cosmology. Handwave And any pre-recombination modification has to survive light-element abundance constraints, which come from nuclear cross-sections measured in laboratories. Both are places where a cosmological claim can be killed by someone who does not read cosmology papers. Neutrino Astronomy and Laboratory Astrophysics carry those measurements.

Frontier Every question in section 4 has a test, and they rank by what a result would settle rather than by how expensive they are. The striking feature of the ranking is that the top item costs almost nothing.

Speculative Tier one, and the decisive experiment for this brief: a blind, jointly specified reanalysis of the same JWST observations by the two competing distance-ladder teams, on a common galaxy sample, with pipelines and cuts registered before unblinding. The two programmes currently report expansion rates that differ by about two standard deviations from distances that agree to about one percent, using the same telescope. Speculative Either the disagreement survives a common analysis, in which case the ladder is measuring something real and the early-universe leg is where the new physics must be; or it does not, in which case the headline tension of the last decade was an analysis choice. Handwave The data exist, the telescope time is spent, and the obstacle is entirely that no institution can compel two collaborations to do it.

Frontier Tier two is the commission’s own decisive experiment, and its numbers need correcting. Standard sirens with electromagnetic counterparts would give an expansion rate with systematics orthogonal to both the ladder and the sound horizon. Speculative The commission puts the requirement at about 50 events; the 2026 review of the method puts it at roughly 100 to 1,000 binary neutron star events with counterparts for 1 to 2% precision. Handwave The delivered count after nine years is one, for reasons Multi-Messenger Astronomy sets out in detail, so this is a third-generation-detector experiment rather than a near-term one.

Established Tier three is the supernova recalibration, and it is the highest ratio of decisiveness to cost anywhere in this brief. Apply one photometric calibration across Pantheon+, Union3 and the Dark Energy Survey sample, by a team that has not published a dark-energy constraint, and rerun the equation-of-state fit. Frontier The 2026 analysis using a recalibrated Dark Energy Survey sample already reports significance of 1.1 to 2.3 standard deviations from late-time data alone. Speculative If a common calibration collapses the 2.8-to-4.2 spread toward its lower end, the evolving-dark-energy result is a zero-point offset; if the spread survives, it is a feature of the universe.

Frontier Tier four attacks the early-universe leg without a satellite. Microwave-background analyses that avoid the sound horizon — using the matter-radiation equality scale or lensing instead — already give 69.07 ± 0.44 from nine measurements, and tightening that class would test whether the early-universe number depends on the one piece of physics every proposed solution modifies. Speculative A sound-horizon-free early-universe value that stays low kills the entire class of pre-recombination solutions at a stroke.

Frontier Tier five settles the clustering question by confrontation rather than by accumulation. Reanalyse the Dark Energy Survey Year 6 and KiDS Legacy weak-lensing data with a single systematics model, single redshift-calibration method and single set of scale cuts. Speculative One is consistent with the microwave background at under one standard deviation and the other is 2.4 to 2.7 standard deviations away. Two surveys of the same sky cannot both be right, and the difference between them is now larger than the tension either claims to measure.

Speculative Tier six is the scheduled accumulation, and it is the one everybody is already doing. Euclid’s first release, Rubin’s survey, DESI’s final dataset and Gaia’s fourth release will each move an error bar. Handwave None of them removes an analysis choice, which is why this tier ranks below three cheap experiments that do.

Speculative Tier seven is a cross-disciplinary check that cosmology cannot run on itself. The cosmological preference for a neutrino mass sum at or below zero is in 3 standard deviation tension with the laboratory lower bound, and it vanishes when the dark-energy equation of state is freed. Speculative A direct laboratory limit tightening below the cosmological preference would force the issue in a way no survey can.

Frontier And the exercises that are not tests and are routinely reported as though they were. Fitting another extension of the standard model to the same combined datasets, re-deriving a significance from a combination chosen after inspecting it, and announcing that a given model resolves the tension are model-selection statements about a fixed dataset. Speculative They produce papers at a rate no observational programme can match and they have never retired a hypothesis, which is the property that distinguishes an experiment from an exercise.

Frontier There is no engineering programme specific to this subject; there is a set of facilities built for survey cosmology whose outputs happen to adjudicate it.

Established The instruments that matter are already built or already funded. A multi-object spectrograph capable of fourteen million redshifts, a space telescope with the infrared resolution to separate individual red giants in a supernova host galaxy, a wide-field imager stable enough to measure shear at the part-in-a-thousand level, and an interferometer network with enough baselines to localise a merger to a few square degrees. Frontier Mega-Telescopes carries the aperture cost scaling and Gravitational Wave Engineering the detector side.

Speculative The one piece of engineering this subject specifically needs is metrological rather than optical. An absolute photometric calibration transferable between telescopes at the millimagnitude level would remove the largest systematic in the supernova case and much of the spread in the ladder. Speculative That is a standards problem of the kind Metrology Infrastructure describes, and it has no owner in astronomy because it produces no papers. A field whose decisive requirement is a traceable calibration chain and which has no metrology institute is in the position optical astronomy was in before photometric standard stars.

Handwave The facility that would most change the picture is the one nobody has proposed seriously enough to cost. A next-generation full-sky microwave-background satellite would re-audit the early-universe leg with independent optics, independent systematics and a different team. Speculative Its absence means the number on one side of the field’s central disagreement has not been independently remeasured at full sky since 2018, and that is an engineering fact rather than a scientific one.

Established The nearest brief on this map is Multi-Messenger Astronomy, and the two pages divide the subject cleanly. That brief is about astronomy as instrument and destination: what a coincident detection costs to obtain, why one neutron-star merger with a counterpart in nine years is a constraint about sky area and spectrograph queues rather than about detectors, and what the single 2017 event settled. Frontier This page is about whether the standard cosmological model is breaking, and it treats the standard siren only as one of half a dozen methods that have failed to referee the expansion-rate disagreement. The counterpart drought is that brief’s finding and this brief’s bottleneck.

Frontier Three adjacencies would inherit a positive result directly. Quantum Gravity and Hidden Sector Searches both contain candidate mechanisms for a pre-recombination modification, and a confirmed tension would be the first observational constraint either has had from cosmology in decades. Speculative Post-LHC Colliders inherits the neutrino-mass anomaly, because a cosmological preference below the oscillation floor is a statement about particle physics made by astronomers.

Frontier Two adjacencies are methodological and matter more than they look. Objective Collapse Models shares this subject’s structural problem exactly: a theory space that acquires free parameters at least as fast as measurement removes them, and no agreed convention for deciding which extensions are live. Speculative Philosophy of Science owns the question of when an anomaly becomes a refutation, which is the question this page cannot answer from inside.

Speculative One adjacency is about how the field knows what it knows. Research Front Detection and Scientific Funding Models both bear on why this literature produces hundreds of model extensions a year and almost no recalibration work. Handwave The distribution of effort here is not a scientific judgement about where the answer lies; it is a reward structure, and section 10 says so.

Frontier Cosmological tensions are produced and adjudicated by a small number of large collaborations, and the institutional problem is that no body sits above them.

Established The collaborations are well run and they blind their analyses. The 2025 time-delay result unblinded on a stated date after the analysis was fixed; weak-lensing surveys blind shear and photometric-redshift calibrations; the DESI collaboration reported performing extra tests specifically to check that its dark-energy result was not driven by an unknown effect in the data. Frontier Blinding within a collaboration does not touch the choice that matters, which is which external datasets to combine with, and that choice is made afterwards.

Speculative The two-team problem is the field’s central institutional failure and it has a cheap fix nobody can impose. Two distance-ladder programmes with different principal investigators analyse overlapping JWST data and report expansion rates two standard deviations apart. Handwave In a discipline with an arbiter, this would be settled by a jointly designed blind reanalysis within a year; in astronomy there is no arbiter, the telescope time allocation committee has no mandate to require it, and the funding agencies fund proposals rather than adjudications.

Frontier Nobody funds the calibration. The single most decisive cheap experiment in section 7 is a uniform photometric recalibration across supernova samples, which produces no new science, threatens published significances, and would be done by people whose careers are measured in discoveries. Speculative The same asymmetry runs through this corpus: the work that most reduces uncertainty is the work least rewarded, and it is visible here as a literature of model extensions sitting on top of a calibration nobody owns.

Speculative One institutional requirement is a convention rather than a facility. There is no agreed standard in this field for how to report the significance of a tension: whether to correct for the number of comparisons made, how to treat correlations between reanalyses of shared data, or what counts as an independent measurement. Handwave Until that exists, the headline number of the whole subject is a quantity computed differently by every group that quotes it, and the 6.5-versus-3.9 standard deviation gap inside a single 2026 paper is what that looks like in practice.

Frontier The ethical questions here are about public knowledge rather than about harm, which does not make them small.

Established The communication asymmetry is measurable. A 3.1 standard deviation preference for evolving dark energy was reported worldwide as a discovery about the nature of the universe; the same year’s reanalyses finding 1.1 to 2.3 standard deviations, the six-year survey that fit the standard model just as well, and the January 2026 audit relocating the expansion-rate problem to the distance ladder received a small fraction of that attention. Speculative A public that was told dark energy is weakening has not been told that the evidence weakened, and the asymmetry is structural rather than dishonest.

Speculative There is a real cost question and it deserves to be stated without pretending to answer it. The facilities that address these questions are national-scale instruments, and the argument for them is that the fundamental nature of the universe is worth knowing. Handwave That argument is strong and it is not improved by overstating how close the answer is; a field that has announced the imminent resolution of the same tension for eight consecutive years is spending credibility it will need.

Frontier The third question is about who gets to be a measurement. The January 2026 audit’s conclusion depends on how 88 published determinations were sorted into categories, and its authors say so. Speculative Meta-analysis in a field with no registry of measurements is a form of authorship over other people’s results, and cosmology has no convention covering it, in contrast to the trial registries that medicine built for exactly this problem.

Speculative And a smaller point that recurs across this corpus. Cosmological anomalies are regularly recruited to support claims about the fine-tuning of the universe, the multiverse, or the necessity of a creator. Handwave Nothing in this brief bears on any of those, and a reader arriving from that direction should notice that the live candidate explanations are a photometric zero-point and a stellar calibration.

Frontier The civilizational implications of this subject are almost entirely about knowledge rather than capability, and they are worth stating precisely for that reason.

Speculative If the dark-energy result holds, the expected long-term future of the universe changes. A cosmological constant gives eternal exponential expansion; an equation of state that crosses w = −1 and is currently above it permits futures in which acceleration slows, stops or reverses, and the fitted quadrant — w₀ above −1 with w_a below zero — is the one in which dark energy weakens with time. Handwave This has no consequence for anything a civilization does on any timescale it can plan for, and it is the single most reported consequence. Civilization Timelines carries the horizons on which such statements start to mean something.

Frontier The consequence that is real is methodological and it generalises. A six-parameter model fitted an enormous body of independent observations for twenty-five years, and the first credible sign of strain turned out, on inspection, to be entangled with the calibration of a three-rung distance ladder and the zero-points of three supernova photometric systems. Speculative Any civilization-scale inference drawn from a small number of precisely quoted numbers inherits this structure: the model is usually fine and the calibration is usually the story. Scientific Revolutions describes the same pattern from the other end.

Speculative The third implication is institutional and it is the one this brief would press. The cheapest decisive experiments in section 7 are all adjudications — a joint blind reanalysis, a common calibration, a shared systematics model — and none of them has an owner. Handwave A civilization that builds instruments faster than it builds ways of settling arguments between the people who use them will accumulate anomalies it cannot retire, and cosmology in 2026 is the clearest available example.

These horizons track data releases and adjudications rather than technologies, because no technology in this subject is missing:

• 10 yr: Frontier Expect Euclid, Rubin, the final DESI dataset and Gaia DR4 to tighten every error bar in section 3 without removing any analysis choice. Speculative Expect the clustering tension to be resolved as survey systematics rather than as physics, on the evidence that it has already changed which survey reports it. Speculative Expect the dark-energy preference to settle between two and three standard deviations and to stop moving, which is the least satisfying and most likely outcome.
• 25 yr: Speculative This is the window in which a third-generation gravitational-wave network could deliver enough counterpart events for a percent-level expansion rate with orthogonal systematics, if the counterpart supply problem is solved rather than merely waited out. Speculative It is also the window in which a next-generation microwave-background satellite could re-audit the early-universe leg, if one is funded, which it currently is not. Handwave A confirmed tension at that point would be the first observational entry point to physics beyond the standard cosmological model since 1998.
• 50 yr: Handwave On this horizon the plausible outcomes are consolidation or dissolution: either the anomaly has become a parameter in a successor model, or the distance ladder was recalibrated and the whole episode is a chapter about systematics. Speculative The dissolution outcome is the modal one and the field is poorly set up to recognise it, because no mechanism exists for declaring an anomaly retired.
• 100 / 250+ yr: Handwave Beyond useful forecasting. Handwave The defensible statement is about shape: a discrepancy that is a fixed percentage rather than a growing one, sitting between a measurement and a model inference, either gets absorbed by a better calibration or becomes the definition of a new parameter, and both have happened repeatedly in the history of this exact quantity.

• Depends on Three briefs on this map own results this page waits on. Multi-Messenger Astronomy owns the counterpart-rate constraint that bounds the standard-siren route, and its finding that one merger with a confirmed counterpart has been found in nine years is the reason section 7 ranks sirens second rather than first. Mega-Telescopes owns the aperture cost scaling that decides how fast the spectroscopic follow-up bottleneck can be relieved. Gravitational Wave Engineering owns the detector sensitivity and network geometry behind localisation. No brief here produces the calibration work this subject most needs, and that absence is the point of the tokens below.
• Requires (not on this map) Six conditions, in the order of the tokens above. The joint reanalysis. Two distance-ladder teams analysing the same JWST galaxies under a pre-registered common pipeline; the data exist and no institution can require it. The calibration. One photometric system applied across Pantheon+, Union3 and the Dark Energy Survey sample by a team with no dark-energy result to defend, which is what the 2.8-to-4.2 standard deviation spread is made of. The satellite. An independent full-sky microwave-background mission; the early-universe leg of the field’s central disagreement has not been remeasured at full sky since 2018 and no successor is funded. The orthogonal measurement. A 1% expansion rate from a method using neither ladder nor sound horizon; the best current candidates are strong-lensing time delays at 4.6% and standard sirens at 6.4%, and both are consistent with everything. The registry. A record of cosmological determinations with stated dependencies, so that a meta-analysis of 88 measurements does not have to invent its own categorisation and then report that the answer is sensitive to it. The convention. An agreed rule for computing and reporting tension significance, including trials factors and correlations — the cheapest item on this list, and the one whose absence lets a single paper quote 6.5 and 3.9 standard deviations for the same comparison.
• Enables A confirmed tension would give Quantum Gravity and Hidden Sector Searches their first cosmological constraint in decades, and would hand Post-LHC Colliders a neutrino-sector target set by astronomy. A resolved tension enables nothing and closes a great deal, which is the more likely outcome and the less discussed one. No typed enabling edge is claimed, because a disagreement about error bars enables nothing on a schedule.
• Adjacent Survey cosmology, stellar astrophysics, supernova photometry and gravitational-wave astronomy supply every number on this page. Within this map: Multi-Messenger Astronomy for the siren route, Neutrino Astronomy for the mass-sum anomaly, Laboratory Astrophysics for the nuclear cross-sections behind light-element constraints, Objective Collapse Models for the shared problem of a parameter space that grows under pressure, Metrology Infrastructure for the calibration chain this subject lacks, and Philosophy of Science for when an anomaly becomes a refutation.

Established Claim in circulation: the Hubble tension is a five-sigma disagreement between two measurements. The arithmetic is right for the 2022 pairing and the sentence has two defects. Established The early-universe number is not a measurement of the expansion rate; it is the value the standard model needs to reproduce the observed acoustic scale, and it moves if the physics before recombination moves. Frontier And there are three audited local programmes, not one: the Chicago–Carnegie results run from 67.80 to 70.39 km s⁻¹ Mpc⁻¹ and their authors say no new physics is required. Speculative The honest sentence is that the local ladder disagrees with everything else at somewhere between 3.9 and 6.5 standard deviations depending on how correlations are handled, and that two teams using the same telescope disagree with each other.

Established Claim: DESI has discovered that dark energy is weakening. What DESI DR2 reported is a 3.1 standard deviation preference, in a combined fit with the microwave background, for a two-parameter equation of state over a constant one. Frontier Adding supernovae moves that to between 2.8 and 4.2 depending on which supernova sample is used, and the spread between samples that share objects is itself the size of the effect. Speculative A 2026 reanalysis using a recalibrated supernova sample reports 1.1 to 2.3 standard deviations from late-time data alone, and a 2025 critique finds that no individual dataset detects acceleration significantly once the equation of state is freed. Handwave Three standard deviations in a combined fit is a hint. It is not a discovery, and the collaboration did not call it one.

Frontier Claim: the S₈ tension shows that structure grew more slowly than the standard model predicts. Established Against a 2026 combined microwave-background baseline of 0.836 ^+0.012_−0.013, the KiDS Legacy analysis is consistent at under one standard deviation and Dark Energy Survey Year 6 is 2.4 to 2.7 standard deviations low. Speculative The survey that supplied this tension for most of a decade no longer reports it, and a different survey does. Handwave A tension that changes instruments while keeping its size is a statement about instruments until somebody shows otherwise.

Speculative Claim: new physics is required. Frontier Every proposed pre-recombination modification has to raise the local expansion rate without spoiling the acoustic peaks, the light-element abundances or the clustering amplitude, and the clustering constraint has tightened against exactly the class of solutions that was most popular. Speculative Meanwhile the live alternative — that the local calibration chain has an unrecognised systematic — has gained rather than lost ground since 2024, because the second JWST programme exists and disagrees. Handwave A reader who is told the tension requires new physics is being told the conclusion of one side of an unsettled argument.

Frontier Claim: gravitational-wave standard sirens will settle this shortly. Established One merger with a confirmed electromagnetic counterpart has been found in nine years, and the 2026 review of the method puts the requirement for 1 to 2% precision at roughly 100 to 1,000 such events. Frontier The two siren analyses published in 2026 give 67.94 ^+4.40_−4.34 from 142 events combined with weak lensing and 78.2 ^+12.0_−11.0 from 17 dark sirens. Speculative Two methods that disagree with each other by more than the tension they are meant to arbitrate cannot yet arbitrate it. Multi-Messenger Astronomy explains why the counterpart supply is the binding term.

Speculative Claim: cosmology is in crisis. Established The competing distance-ladder programmes agree on distances to individual galaxies at about the 1% level; the microwave-background, oscillation-scale and light-element measurements remain mutually consistent to high precision; and the standard model still fits a very large body of data with six parameters. Speculative What is in dispute is a small number of derived quantities, and the word for that is an anomaly rather than a crisis. Handwave The rhetorical inflation is not harmless: it is what makes a 3.1 standard deviation preference reportable as a discovery.

Speculative Claim: cosmology has measured a negative neutrino mass. Frontier What exists is a preference, in a fit of the standard model to the combined data, for a mass sum at or below the laboratory lower bound, at about 3 standard deviations, which vanishes when the dark-energy equation of state is allowed to evolve. Speculative A negative value of a quantity that cannot be negative is a diagnostic that something in the fit is wrong, not a measurement of the quantity, and the fact that one extension removes two anomalies at once is as easily read as an indictment of the extension’s flexibility as an endorsement of it.

Speculative Claim: some particular model resolves the Hubble tension. Handwave Papers making this claim appear at a rate no observational programme can match, and none has retired a competitor. Speculative The reason is structural: these are model-selection statements about a fixed body of data, made after the data are known, in a field with no agreed convention for trials factors. The test of a proposal here is not whether it fits, but whether it forbids something an instrument could look for, and very few of them do.