At Full Strength · the wave’s portal

Every Claim at Full Strength

The At Full Strength wave took extraordinary claims from the record of science, from a counting horse to faster-than-light neutrinos. For each one it rebuilt the claim from its authors’ own numbers, then tested the test that decided it (or, for a claim still open, the test that would): by planting the claim into the control’s own data where those were released, and by setting it against the control’s printed sensitivity where they were not. This portal reads every page’s machine record and nothing else. Sort the claims by verdict, by what decided them and by how long deciding took; see every cell of every page’s ledger, the false ones included; and read what the wave would not build, and why.

Each page of the wave did four things, in order. It reproduced the claimants’ own headline evidence from their own published numbers, with its own engine, before anything else. It put the test that historically decided the claim (or, for a claim still open, the test that would) in the reader’s hands: rerun on real data where the control’s data were released, and read from the control’s published summary where they were not, which the ledger wall’s control-computed column records page by page. It tested that test: where the control’s data were released, it planted the claimed effect, at its claimed size, into a copy of them and reran the unmodified control (grade A); where the control was published only as a sensitivity, it worked out the control’s power from its authors’ own printed numbers and said so (grade B). And it stated the verdict in one of five words, with a date, its sources and what would change it.

Every page ships a small record of all of that beside it, and a supervisor gate compares that record’s ledger with what the page’s own verifier computes. This page reads those records and nothing else. It cannot say anything about a claim that the claim’s own page did not already stand behind, and it cannot check a claim itself.

Without JavaScript this page shows the claims as one table, written from the records by the index generator. With it, the page fetches every record itself.

I · The ten-second version

The claims, side by side

Each card is one claim: its title, the year it was made, the headline as its authors printed it, and the verdict with the date it was given. Switch the arrangement and watch the same claims regroup.

Every claim in the index, from each page’s own record.json, as the index generator last wrote it.
Claim	Claimed	Headline, as printed	Verdict	The verdict’s own bounds	As of	Decided by	Years to decision	Power grade
The Anomaly Outlived Its Planet	1859	CRAS 49, p. 380: augmenter de 38 secondes le mouvement séculaire du périhélie; Annales t. V, p. 96: π′ = +0″,383 − 2,72 e′ − 0,001 67 τ − 0″,033 ν + 0,87 δn″; p. 100: (A) 288″ ν′ + 87″ ν″ = 38″,3	VINDICATED	none stated in the record	2026-09-23	independent replication	23	grade A, injection
The Sum Nobody Knew	1905	cards nobody present could see: er schlug beim ersten Zettel 11 und schlug für die erste Zahl 4, für die zweite 7; Ebenso bei drei anderen Zetteln. Tablets: dieser antwortete mit 9 Hufschlägen. Jetzt wurden die Tafeln umgedreht und es zeigte sich, dass 5 und 4 geschrieben war. Beide Versuche wurden wiederholt und gelangen.	ARTEFACT	none stated in the record	2026-09-22	contamination	0	grade A, injection
The Plates That Were Set Aside	1920	Sobral 4-inch: 1".98 with a probable error of about ±0".12; Principe: 1".61, probable error about ±0".30; Sobral astrographic plates, given little weight: 0".93 (1".52 with the refraction and aberration scale). Alternatives printed on p.291: Newtonian 0".87, Einstein 1".75.	VINDICATED	The verdict’s scope, in the record’s words: the claim as the 1920 report makes it: that light passing the Sun is deflected by the amount Einstein’s theory demands, not the Newtonian half A narrower question the record also answers: The set-aside Sobral astrographic plates on the 1979 remeasurement alone: INCONCLUSIVE between Einstein and Newton.	2026-09-22	independent replication	56	grade B, published sensitivity
The Test for Intellectual Blooming	1966	Table 1 (1966), weighted: controls M 8.4 (sigma 13.5), experimentals M 12.2 (sigma 15.0), Diff. 3.8, t 2.15, p .02 one-tailed; book Table 7-1: N 255 gain +8.42, N 65 gain +12.22, advantage +3.80, p .02, mean square within treatments within classrooms 164.24; grade 1 +15.4 (t 2.97, p .002), grade 2 +9.5 (t 2.28, p .02). Text: For the school as a whole those children from whom the teachers had been led to expect greater intellectual gain showed a significantly greater gain in IQ score than did the control children (p = .02, one-tail).	OPEN	none stated in the record	2026-09-23	not yet decided	open	grade A, injection
When the Whole Field Moved	1969	P = 25 years, e = 0.75, T = 1950; semi-axis major 0″0275 (α = 1.46 μ); companion 0.0016 solar masses, 1.7 Jupiter masses	ARTEFACT	none stated in the record	2026-09-22	the instrument	6	grade B, published sensitivity
The Face an Hour Before Sunset	1988	Peak height 412.5 ± 17.5 (4.2%)	ARTEFACT	none stated in the record	2026-09-23	perception	13	grade A, injection
Nine Points for Ten Minutes	1993	mean SAS 57.56 (music), 54.61 (relaxation), 54.00 (silence); IQ equivalents 119, 111, 110; "8–9 points"; F2,35 = 7.08, P = 0.002; Scheffe’s t = 3.41, P = 0.002; t = 3.67, P = 0.0008; t = 0.795, P = 0.432	DISSOLVED	none stated in the record	2026-09-23	independent replication	17	grade A, injection
The Hours Between Best and Good	1993	best violinists 7,410 hr against good violinists 5,301 hr, F(1, 27) = 4.59, p < .05; mean of best and good against the music teachers' 3,420 hr, F(1, 27) = 11.86, p < .01; accumulated practice alone at age 18, 10 violinists a group	OPEN	The verdict’s scope, in the record’s words: whether a conservatory's best student violinists had accumulated more solitary practice by 18 than its good ones (the best-over-good step of the 1993 complete correspondence)	2026-09-22	not yet decided	open	grade A, injection
The Dimming That Turned Around	1998	10% to 15% farther than expected in a low mass density (ΩM = 0.2) Universe without a cosmological constant (abstract); "The average difference between the data and the ΩM = 0.20, ΩΛ = 0 prediction is 0.25 mag." (Figure 4 caption, MLCS); p(q0 ≤ 0) = 99.5% (2.8σ), p(ΩΛ ≥ 0) = 99.7% (3.0σ), MLCS+Snap.(15), Table 8	VINDICATED	none stated in the record	2026-09-23	more data	6	grade A, injection
The Odds Against the Earl	2004	Table 3.1 (p. 370): total Shakespeare rejections 7 (Oxford) against 1; Discrete Comp. Prob. 3000 7.75E-07 against 3.08E-01, "Sh.: Oxford ≈ 400,000 to 1"; Continuous Comp. Prob. 3000 <1E-15 against 1.50E-01, "Sh.: Ox. ≈ 150 trillion to 1". Appendix Six (p. 424): Oxford 6 rejections in 14 tests, discrete 7.749E-07, continuous composite error 11.9588, continuous composite <1.0000E-15	OPEN	none stated in the record	2026-09-23	not yet decided	open	grade A, injection
The Melt That 2035 Would Need	2007	the likelihood of them disappearing by the year 2035 and perhaps sooner is very high; Its total area will likely shrink from the present 500,000 to 100,000 km² by the year 2035	WITHDRAWN	none stated in the record	2026-09-23	the analysis	3	grade A, injection
The Proton That Stayed Small	2010	rₚ = 0.84184(67) fm, differing by 5.0 standard deviations from 0.8768(69) fm.	VINDICATED	The verdict’s scope, in the record’s words: VINDICATED applies to the small-radius scale, independently supported by electronic hydrogen. It does not preserve every digit of the original theory, establish a new force, or settle every scattering analysis.	2026-09-22	independent replication	16	grade A, injection
The Clock That Moved the Particle	2011	delta t = TOF_c - TOF_nu = (60.7 +/- 6.9 (stat.) +/- 7.4 (sys.)) ns, with 6.0 sigma significance (arXiv:1109.4897v1, pages 18-19)	ARTEFACT	none stated in the record	2026-09-22	the instrument	1	grade A, injection
Thirty-Seven Thousand Curtains	2011	53.1%, t(99) = 2.51, p = .01, d = 0.25	DISSOLVED	The verdict’s scope, in the record’s words: An above-chance hit rate on erotic trials in the Experiment 1 paradigm of Bem (2011). Nothing about the other eight experiments of that article, or about ESP in general.	2026-09-22	independent replication	12	grade A, injection
The Dust Between the Stars	2014	r = 0.20 (+0.07, -0.05), with r = 0 disfavored at 7.0 sigma (arXiv:1403.3985v3 abstract; Section XI.1: L(0)/L(peak) = 2.9e-11, PTE 3.3e-12)	ARTEFACT	What the verdict is about, in the record’s words: the attribution of the 2014 excess to primordial gravitational waves near r = 0.2	2026-09-22	contamination	1	grade A, injection
The Visitor Gravity Could Not Fit	2018	A₁ = (4.92 ± 0.16) × 10⁻⁶ m s⁻²; about 30σ	VINDICATED	none stated in the record	2026-09-22	the analysis	0	grade A, injection
Where the Missing Neutrons Could Go	2018	879.6 ± 0.6 s (bottle); 888.0 ± 2.0 s (beam); Δτ_n ≃ 8.4 s; 4.0 σ; Br(n → p + anything) ≈ 99%; approximately 1% dark branch.	OPEN	none stated in the record	2026-09-22	not yet decided	open	grade B, published sensitivity
The Line Beneath the Ripple	2020	quality up to ~15σ (abstract); Table 1 ALMA line-integrated signal-to-noise: mid-latitude 14.5, whole planet 13.3, equator 5.0; ~20 ppb from the authors’ radiative-transfer model (not computed here)	OPEN	none stated in the record	2026-09-22	not yet decided	open	grade A, injection
The Seeds Beneath the Steps	2021	calibrated ages that range from 22.86 ± 0.32 to 21.13 ± 0.25 ka (N=11)	OPEN	none stated in the record	2026-09-22	not yet decided	open	grade B, published sensitivity
The Year That Would Not Disappear	2021	2-6 keV, 2.86 ton×yr: A = (0.01014 ± 0.00074) cpd/kg/keV, 13.7 σ, T = (0.99834 ± 0.00067) yr, t0 = (142.4 ± 4.2) days; fixed T = 1 yr and t0 = 152.5 days: A = (0.00996 ± 0.00074) cpd/kg/keV, 13.4 σ, χ²/d.o.f. = 130/155	OPEN	none stated in the record	2026-09-22	not yet decided	open	grade A, injection

II · A timeline of deciding

From the claim to its decision

Each claim is a bar from the year it was made to the year its record says it was decided, which is the claim’s year plus the record’s own years to decision. A claim still open is drawn as an unfinished bar that runs to the day of its record’s as-of date and stops there, labelled open. It is never given an end.

Draw the bars on the calendar all starting at the claim

III · The ledger wall

Every cell of every ledger

The wave’s CONTRACT gives each page nine yes-or-no cells, each backed by a named assertion in that page’s verifier: did the page reproduce the claim, compute the control, show the control could have confirmed the claim, assert its printed figures, make its choices live, catch a planted fault, assert a refusal, pin its data, and date its verdict. A false cell is the honest thing it is, such as a grade B page recording that its control could not be rerun. None is hidden and none is averaged. There is no score here of any kind, because the CONTRACT forbids one on purpose.

Show assertion ids only on the false cells Show only the pages with a false or missing cell

IV · The second layer

What the records say together

These numbers describe this wave’s selection, not science. The claims were chosen by hand from the record of science, and the wave kept only those whose claimants’ numbers, and whose control’s data or printed sensitivity, were public in the same observable. Nothing below is a sample of extraordinary claims, and nothing below is a base rate for how often they come true. Each figure is computed in your browser from the records above and printed with the count it rests on.

How the claims came out

How long deciding took, by what decided it

Did claims settled by a fault in the instrument or by contamination settle faster than claims that needed someone to repeat the work? Each row is one kind of decision, with every claim’s years to decision and their median.

Looking again at evidence already in hand needs no new experiment, and new evidence has to wait for one, so part of this gap is expected by construction. Each kind also mixes verdicts: the comparison says how deciding happened in this selection, not which claims were right.

Could the control be rerun?

Cross any two

Choose what the rows and columns count, then choose a cell to name the claims in it.

Rows Columns

V · What the wave would not build

The candidates it killed, and why

The wave screened more candidates than it built. A page could be built only when two things were public in a form it could use: the claimants’ own numbers, enough to reproduce their headline at the strength they printed, and the deciding control’s data or sensitivity, in the same observable as the claim, so that the claimed effect could be planted into it, or set against it, at its claimed size. When either half was missing, or what was public could not carry the test, the spec stage killed the candidate and kept the objection. The screen is itself a finding: each kill names the half of the evidence on which the test broke down, whether because the numbers are not public or because what is public does not measure the claim at its claimed size.

Which side a kill fell on is this page’s own reading of the quoted objection, not the wave’s, and each objection is printed beside it so you can disagree. The claim side means the claimants’ headline could not be reproduced at the strength they printed from what they published; the control side means the claimed effect could not be planted into the deciding control’s own observable, or set against its published sensitivity, at its claimed size. Each objection is quoted word for word from the candidate’s own file in the wave’s directory, which holds the full argument.

Show the kills that fell on either side the claim side the control side both

First pass, under the strict power rule (before grade B existed)

LK-99 lk99 the control side
“A large resistance drop is already present in the non-superconducting Cu2S control. Recovering another planted drop therefore cannot show that the control could have confirmed the superconductivity claim.”
specs/KILLED-lk99.md
the Theta+ pentaquark pentaquark-theta the control side
“The prospect's proposed fallback power test does not plant the claimed physical effect. It transfers 19 detected LEPS events into a different experiment without a justified conversion of production rate, exposure, acceptance, and nuclear final-state interactions.”
specs/KILLED-pentaquark-theta.md
the 750 GeV diphoton excess diphoton-750 both sides
“The claimed signal yield is a parameter of a separate signal-plus-background fit, with the background parameters fitted again. A residual above the first curve is not the signal yield from the second fit.”
specs/KILLED-diphoton-750.md
polywater polywater the control side
“The proposed injection changes a spectrum, but the decisive question is chemical identity.”
specs/KILLED-polywater.md
the fifth force read out of Eötvös fifth-force-eotvos the control side
“the suggested control's own exclusion figure has contradictory sign labels, and the curve labelled repulsive is clipped at the claimed range.”
specs/KILLED-fifth-force-eotvos.md
the neutron lifetime puzzle neutron-lifetime the control side later specified again and built as where-the-missing-neutrons-could-go
“Comparing an added line with a published exclusion curve checks the published answer, not whether the control detects the planted effect.”
specs/KILLED-neutron-lifetime.md
the X17 anomaly x17-atomki the control side
“MEG II uses a likelihood with finite simulation statistics and nuisance parameters. Its displayed central curves do not supply the effective simulation counts, separate component templates, or their changes with mass and calibration.”
specs/KILLED-x17-atomki.md
the Wow! signal wow-signal the control side
“The proposed replacement power test changes the physical signal while claiming to plant the historical one.”
specs/KILLED-wow-signal.md
the Allais effect allais-effect the control side
“The public Salva figures contain corrected precession rates. They do not contain the uncorrected azimuth and ellipse measurements required to rerun his control.”
specs/KILLED-allais-effect.md

After the grade B rule was added

Planet Nine planet-nine-clustering the control side
“the published 0.007% is a tail probability for an already selected sample, not a specified intrinsic population effect that can be transplanted into a different survey at the claimed size.”
specs/KILLED-planet-nine-clustering.md
cold fusion cold-fusion-gamma the claim side
“The proposed page combines a transformed gamma plot with a neutron-rate headline from a different measurement.”
specs/KILLED-cold-fusion-gamma.md
the fifth force read out of Eötvös fifth-force-eotvos the claim side
“The reproducible EPF slope does not determine the coupling and range against which the proposed grade B control would be compared.”
specs/KILLED-fifth-force-eotvos.md
the X17 anomaly x17-atomki the claim side
“The checked claimant record does not specify the actual null chi-square, its degrees of freedom, the exact bins entering that significance, or a complete treatment of fitted background and shared acceptance uncertainty.”
specs/KILLED-x17-atomki.md
L'Anse aux Meadows vinland-1021 the claim side
“The earliest complete six-ring fit available from that reference is consequently 1021. A minimum at this boundary cannot establish the reported preference over 1019 or 1020.”
specs/KILLED-vinland-1021.md
methane on Mars mars-methane the control side
“the pressure-normalised TLS concentration does not specify the methane column that a TGO occultation would encounter”
specs/KILLED-mars-methane.md
solar neutrino problem solar-neutrino-problem the claim side
“The published run estimates and the headline fit are outputs of different statistical procedures.”
specs/KILLED-solar-neutrino-problem.md
Stonehenge's Altar Stone altar-stone the control side
“The proposed power injection checks whether a difference test tolerates a manufactured match. It does not establish sensitivity to the claimed Scottish provenance.”
specs/KILLED-altar-stone.md
the peppered moth peppered-moth the control side
“The proposed instrument cannot test bird-specific predation at the historical claimed size.”
specs/KILLED-peppered-moth.md
the Draupner wave draupner-wave the control side
“the laboratory result is conditioned on the historical waveform: the experiment adjusts the wavemaker amplitudes and phases towards the Draupner measurement.”
specs/KILLED-draupner-wave.md
seasonal radioactive decay decay-rate-seasonal the claim side
“The available claimant record does not support a checked reproduction of the headline correlations at their printed strength.”
specs/KILLED-decay-rate-seasonal.md

Killed, and listed by the slate only among the pages later built

OPERA's superluminal neutrinos opera-ftl the claim side later specified again and built as the-clock-that-moved-the-particle
“the published likelihood figure is not a demonstrated numerical substitute for the missing inputs, even when its fitted curves are extracted directly”
specs/KILLED-opera-ftl.md

VI · The check

What this page did, and what it cannot do

What it read

The index, claims.json, lists every record.json on this site whose top level has a claim object and a verdict object, with the SHA-256 of its bytes. The page fetches the index, then fetches every record it lists from this site and hashes the bytes it received against the index.

The CONTRACT’s own rules, checked against every record

What it refuses

The timeline refuses to draw a record for any of these reasons, and names each refused record below with its reason: the record gives no whole-number claim year; years to decision is not a whole number of years from zero up; the verdict carries no valid as-of date; an OPEN verdict that also gives a decision year; a settled verdict with no years to decision; a decision year after the record’s own as-of date; a claim year after the record’s own as-of date.

Every free choice it makes

A decision year is the claim’s year plus the record’s own years to decision. The page never dates a decision itself, and where a record says how it counted its years, the card quotes it.
On the timeline a year is drawn from its first day, and an open claim runs to the first moment of the day of its record’s as-of date.
Fields are grouped on the record’s field text up to any parenthesis, lower-cased, so that psychology pages with different subtitles in brackets share a group.
A median is the middle value, or the mean of the two middle values.
For one comparison the page pairs the CONTRACT’s words for what decided a claim into two kinds: looking again at the original evidence (the instrument, the analysis, contamination, perception) and new evidence (more data, independent replication). The pairing is this page’s.
“Settled against the claimants” means DISSOLVED, ARTEFACT or WITHDRAWN.
The side each kill fell on is this page’s reading, printed beside the objection it reads.

What it cannot do

It cannot check any claim. A record is exactly as good as its page’s verifier. What stands between the two is the supervisor’s gate, which runs each page’s verifier and refuses a record whose nine cells disagree with what the verifier computes.
It does not judge between pages. It counts, groups and draws; it never adds cells together, never ranks, and prints no score.
Its counts describe the claims this wave chose and could build. The kills are part of that choice, which is why they are printed.