Events

Events: the record, week by week

One dated record of what the engine read, what was released and what others published, in the week each line was listed. A reading names its record, the record's own word, the verdict as the statuses record maps it and its status kind; an outside line names its source and terms. A reading that failed or was refuted comes before one that held.

How the page is made

The page is generated from one committed record. A reading is published only if the record it names is in the tree and says the verdict printed, at the fragment named; its status kind and date equal the statuses record's; a number in its words is a number in the record; a count equals the tree's; and the datasets it reads carry their credit lines from the data register. An outside line is published only with its url, citation, terms and the date it was read. A line that fails any of these is refused by name and never edited.

Verdicts used on this page:

falsified. The registered prediction did not come out as registered.
refuted. The data refuted the registered prediction.
window-fragile. The result holds under some windows and not under others, so it depends on the window chosen.
held-at-threshold. It came out as registered at exactly the count it needed, with no margin.
informative against random windows only. The test separates the result from windows placed at random. It does not separate the model from the textbook account, so it is not counted as support.
held. The registered prediction came out as registered.
none. A fixed rule was read from public data. Nothing was predicted, so nothing held or failed.

Status kinds used on this page:

measured. The numbers come from recorded observations.
measured-in-model. It was measured on planted or simulated data, inside the model, not on nature.

Week of 11 October 2026

11 to 17 October 2026

Readings

2026-10-04 · reading · falsified The direction the trade desk records is not the direction of the series. We registered that the direction the trade desk's writer records equals the sign of the series' move in at least 0.9 of the series. It matched in 12 of 24 series, and every row read the same direction. Limit: Run on committed series, not on a live account. Record: deploy/account_direction.json (verdicts/F-A1.b). The record's word: falsified. Verdict: falsified. Status: measured-in-model.
2026-10-04 · reading · falsified The clock-rate prediction failed on a planted record. We registered that the energy gap between a pair's two states sets the rate of its clock. On a planted record the correlation between the account's rate imbalance and the clock rate was -0.063, where at least +0.5 was required, and 0 of 20 beats with equal rates had the clock rate vanish. Limit: Run on a planted record inside the model, not on nature. Record: deploy/theta_clock.json (verdicts/F12). The record's word: falsified. Verdict: falsified. Status: measured-in-model.
2026-10-04 · reading · falsified A fixed order of steps is not as good as a random order. We registered that the protocol's fixed order of steps gains no more information than a random order, within a band registered around 1. On the test the ratio was 0.891: the fixed order learned more slowly than a random one. Limit: Run on our own fixtures inside the model; it tests the implementation, not nature. Record: deploy/falsify_sim.json (verdicts/S5.b). The record's word: falsified. Verdict: falsified. Status: measured-in-model.
2026-10-04 · reading · falsified A broker's centrality did not follow the model's m. We registered that, on planted stars, a broker's centrality reads as the size of the model's m in at least 95 percent of cases. It matched in 22 of 120 planted stars, all of them stars with no rim. Limit: Planted stars inside the model, not nature. Record: deploy/falsify_sim.json (verdicts/S12). The record's word: falsified. Verdict: falsified. Status: measured-in-model.
2026-10-04 · reading · falsified The two lock conditions are not one condition. We registered that the emission condition, which compares entropies, and the alignment condition, which compares the heads' directions, agree on at least 0.9 of beats. On 200 planted beats they agreed on 63, a fraction of 0.315, so the claim that they are one condition failed. Limit: Planted beats inside the model, not nature; the two conditions now carry two names. Record: deploy/theta_clock.json (verdicts/c). The record's word: falsified. Verdict: falsified. Status: measured-in-model.
2026-10-04 · reading · falsified A registered lens rule failed on the Earth stand-in. A registered rule for our analysis lens on a stand-in for the Earth's record failed: the fitting step ran a parameter to the edge of its allowed range, and the confidence of the blocks left the interval the rule allowed. Limit: A stand-in record inside the model, not the Earth's own. Record: deploy/lens_cards.json (verdicts/earth_standin/R6). The record's word: FALSIFIED. Verdict: falsified. Status: measured-in-model.
2026-10-04 · reading · falsified Two registered lens rules failed on a planted trading record. Our analysis lens failed two of its registered rules on a planted trading record: it did not reach the registered read-back rate at the top amplitude, and its null was read back within the noise of the planted rate. Limit: A planted record inside the model, not a market. Record: deploy/lens_cards.json (verdicts/trading/R4, verdicts/trading/R5). The record's word: FALSIFIED. Verdict: falsified. Status: measured-in-model.
2026-10-07 · reading · falsified A recorded count did not re-derive. A count recorded in an earlier document, the number of coupled pairs a condensate sees after reweighting, did not re-derive on the same inputs. The direction of the finding holds; the number has been re-recorded. Limit: Which of two changes moved the number has not been attributed. Record: deploy/hilbert_experiments.json (rows/key=HILB-2-CONDENSE-KEPT). The record's word: falsified. Verdict: falsified. Status: measured-in-model.
2026-10-07 · reading · falsified Fourteen registered ledger tests failed as registered. Fourteen tests registered from the papers' own evidence ledgers failed as registered, each on its own numbers. The records name every one. Limit: Each ran on our own fixtures or engine, not on nature; where a test failed, the paper's sentence it tested is what is in question. Record: fixtures/paper_experiments/ctg_f7_feasible_flag.json, fixtures/paper_experiments/sra_f2_nonstored_cues.json, fixtures/paper_experiments/sra_f3_settling_ledger.json, fixtures/paper_experiments/sra_f9_hash_chain_rewrite.json, fixtures/paper_experiments/sra_f5_noise_seeds.json, fixtures/paper_experiments/oua_f1_entropy_reward_squad.json, fixtures/paper_experiments/oua_f3_q_learning_churn.json, fixtures/paper_experiments/oua_f4_misspecified_posterior.json, fixtures/paper_experiments/oua_f7_core_vs_certificate.json, fixtures/paper_experiments/oua_f8_myerson_satterthwaite.json, fixtures/paper_experiments/halo_f2_separable_witness.json, fixtures/paper_experiments/halo_f8_coverage.json, fixtures/paper_experiments/ext_f11_hyperbit_address.json, fixtures/paper_experiments/sra_f6_variational_learner.json. The record's word: falsified. Verdict: falsified. Status: measured-in-model.
2026-10-07 · reading · held The Earth's dipole-moment record is bimodal, as registered. We registered, from the Earth paper's evidence ledger, that the signed dipole-moment record is bimodal. A dip test on the public record read it as bimodal, so the claim held. Limit: It is one public record read by the registered rule: it says the record is bimodal, not why. Record: fixtures/paper_experiments/ert_f1_dip_test.json. The record's word: held. Verdict: held. Status: measured. Credits: EarthRef PADM2M (Ziegler, Constable, Johnson and Tauxe 2011), used for research under the source's published terms; written confirmation pending. Licence: the publisher's published terms. GTS2020 polarity time scale, Ogg 2020, via PmagPy (BSD-3-Clause). Licence: BSD-3-Clause.
2026-10-07 · reading · held The model library's census re-derived. The library holds 99 separate model spaces in three cohorts that share no input: 77 vision, 21 text and 1 audio. The census re-derived within its stated tolerance. Limit: A count of what the library holds, not a measure of what it can do. Record: deploy/hilbert_experiments.json (rows/key=LIB-CENSUS-SPACES). The record's word: held. Verdict: held. Status: measured-in-model.
2026-10-07 · reading · held Four registered ledger tests held. Four tests registered from the papers' own evidence ledgers held on their own numbers, among them the Hebbian capacity law and the path-independence of the discount regime. Limit: All four ran on our own fixtures or engine, not on nature; one further test could not be decided as registered. Record: fixtures/paper_experiments/ctg_f10_furnace_patterns.json, fixtures/paper_experiments/sra_f1_hebbian_capacity.json, fixtures/paper_experiments/sra_f12_read_time.json, fixtures/paper_experiments/oua_f2_discount_regime.json. The record's word: held. Verdict: held. Status: measured-in-model.

Releases

2026-10-13 · release The launch. The Mission pages go out laid out around the objects the engine models, each with its status word, together with the atlas of figures drawn from the records and this page. The press line is pending. Limit: The press line is not stated here: it is pending, and nothing on this page is a press statement.

Week of 4 October 2026

4 to 10 October 2026

Elsewhere

2026-10-04 · external World Space Week begins. World Space Week runs from 4 to 10 October every year. The United Nations General Assembly declared it in resolution A/RES/54/68. Limit: This line cites the dates and the resolution only; the 2026 theme is not cited because the week's own theme page could not be read on 7 October 2026. Source: United Nations General Assembly resolution A/RES/54/68; worldspaceweek.org. Terms: A date range and a resolution number are cited; no text of either is reproduced. Read on 7 October 2026.

Readings

2026-10-04 · reading · falsified A seismic control failed: a registered null did not hold. Our magnetometer experiment around earthquakes registered a control, the Mineral earthquake read at one-minute resolution, that should have read as no signal. It read as a signal that exceeded its bound, so the registered null failed for that record. Limit: No magnetic-field claim from this experiment can be attributed to an earthquake without a remote reference. Record: fixtures/seismo/seismo_lens_results_2026_10_04.json (calls/5). The record's word: fails. Verdict: falsified. Status: measured. Credits: INTERMAGNET 1-minute data, used for research under the source's published terms (non-commercial); no value of it is shown. Licence: CC-BY-NC-4.0. IRIS/EarthScope. Used for research under the source's published terms; written confirmation is pending. Licence: the publisher's published terms.
2026-10-04 · reading · falsified A second seismic call failed: the expected apparent signal did not appear. For the Indios earthquake we registered that the three-component magnetometer would show a large apparent signal that our motion rule would reject by name. No component exceeded its bound, though the motion rule did fire, so the call failed as registered. The same call held for the Sand Point earthquake. Limit: Two earthquakes are too few to say whether the motion rule works. Record: fixtures/seismo/seismo_lens_results_2026_10_04.json (calls/4). The record's word: fails. Verdict: falsified. Status: measured. Credits: USGS Geomagnetism Program (public domain). Licence: public-domain-US-Gov. IRIS/EarthScope. Used for research under the source's published terms; written confirmation is pending. Licence: the publisher's published terms.
2026-10-04 · reading · refuted The Earth's orbital-period test was refuted. We registered, before the data was read, that three orbital periods (23, 41 and 100 thousand years) would all stand out in the Earth's dipole-moment record. Two of the three did. The 23-thousand-year period did not, so the prediction is refuted. Limit: The 41-thousand-year band is usually read as climatic contamination of such records, and this test does not separate that reading from ours. Record: fixtures/earth/earth_readings_2026_10_04.json (s8_orbital_periods). The record's word: holds=false. Verdict: refuted. Status: measured. Credit: EarthRef PADM2M (Ziegler, Constable, Johnson and Tauxe 2011), used for research under the source's published terms; written confirmation pending. Licence: the publisher's published terms.
2026-10-07 · reading · window-fragile The Earth's reversal-window result did not survive a narrower window. We registered, before running it, a check of whether that result survives other windows. It does not. Halving the window leaves 6 of 9 boundaries met, and so does moving every boundary 15 thousand years younger. Widening the window or moving the boundaries 15 thousand years older leaves 7 of 9. Limit: The first result stands as registered; this check reads the same records under other windows and does not change its verdict. Record: deploy/earth_p1_sensitivity.json (verdicts/A_window). The record's word: window-fragile. Verdict: window-fragile. Status: measured. Credits: EarthRef PADM2M (Ziegler, Constable, Johnson and Tauxe 2011), used for research under the source's published terms; written confirmation pending. Licence: the publisher's published terms. GTS2020 polarity time scale, Ogg 2020, via PmagPy (BSD-3-Clause). Licence: BSD-3-Clause.
2026-10-04 · reading · held-at-threshold The Earth's reversal-window prediction held, exactly at its threshold. We registered that the Earth's dipole moment sits at or below its record's tenth percentile within 20 thousand years of at least 7 of the 9 polarity boundaries. It did, at 7 of 9, the least the prediction allowed. Limit: It held with no margin, and boundaries older than the record could not be tested; read with the two lines dated 7 October it is held at its threshold; window-fragile; consistent with the textbook expectation and not as support. Record: fixtures/earth/earth_readings_2026_10_04.json (prediction). The record's word: holds=true (met 7 = need 7). Verdict: held-at-threshold. Status: measured. Credits: EarthRef PADM2M (Ziegler, Constable, Johnson and Tauxe 2011), used for research under the source's published terms; written confirmation pending. Licence: the publisher's published terms. GTS2020 polarity time scale, Ogg 2020, via PmagPy (BSD-3-Clause). Licence: BSD-3-Clause.
2026-10-07 · reading · informative against random windows only Random windows rarely pass the Earth's reversal-window test. We also asked how often windows placed at random meet the same rule. 19 of 1,000 random draws of nine windows did, a rate of 0.019, below the 0.05 level registered for it. The rule separates the Earth's result from random windows. Limit: That is the test's power against random windows only: a weaker field at every reversal is what the textbook expects, so the result is consistent with the textbook expectation and not counted as support. Record: deploy/earth_p1_sensitivity.json (verdicts/B_null). The record's word: informative. Verdict: informative against random windows only. Status: measured. Credits: EarthRef PADM2M (Ziegler, Constable, Johnson and Tauxe 2011), used for research under the source's published terms; written confirmation pending. Licence: the publisher's published terms. GTS2020 polarity time scale, Ogg 2020, via PmagPy (BSD-3-Clause). Licence: BSD-3-Clause.
2026-10-04 · reading · held The analysis lens passed all six rules on the Planck spectrum. We ran our analysis lens on the same published Planck temperature spectrum under six rules registered before the run. All six held. Limit: This calibrates the lens against the published best fit; it is a check on the instrument and tests nothing about our model. Record: deploy/lens_cards.json (verdicts/planck_tt). The record's word: held. Verdict: held. Status: measured. Credit: Based on observations obtained with Planck (http://www.esa.int/Planck), an ESA science mission with instruments and contributions directly funded by ESA Member States, NASA, and Canada. Planck Collaboration 2020, 'Planck 2018 results. VI. Cosmological parameters', A&A 641, A6. Used for research under the source's published terms; written confirmation is pending. Licence: the publisher's published terms.
2026-10-04 · reading · held Two registered seismic nulls held. For the Indios and Sand Point earthquakes we registered that the field-strength witness would not exceed its bound between the arrival of the P and S waves. In both records it stayed under its bound. Limit: In both records the later window could not be calibrated and no distant reference was subtracted; two further calls could not be measured and one could not be attributed. Record: fixtures/seismo/seismo_lens_results_2026_10_04.json (calls/0, calls/1). The record's word: holds. Verdict: held. Status: measured. Credits: USGS Geomagnetism Program (public domain). Licence: public-domain-US-Gov. IRIS/EarthScope. Used for research under the source's published terms; written confirmation is pending. Licence: the publisher's published terms.
2026-10-04 · reading · none The Sun's sound, read from published p-mode frequencies. Reading the published low-degree p-mode frequencies of the Birmingham Solar Oscillations Network by a fixed rule, the large separation, the spacing between consecutive modes, is 135.317 µHz with an uncertainty of 0.060 µHz. Limit: A reading of public measurements by a fixed rule, with no prediction behind it: it tests nothing about our model. Record: fixtures/spectra/sun_2026_10_04.json. The record's word: basis: public_data (a reading; no registered verdict). Verdict: none. Status: measured. Credit: BiSON, University of Birmingham, CC BY 4.0. Licence: CC-BY-4.0.
2026-10-04 · reading · none The Earth's gravest free oscillation, from a published table. The Earth rings after a great earthquake. Its slowest observed mode, called 0S2, has a frequency of 309.45 µHz in the published table of Masters and Widmer. Limit: A reading of a published table by a fixed rule, with no prediction behind it: it tests nothing about our model. Record: fixtures/spectra/earth_2026_10_04.json. The record's word: basis: public_data (a reading; no registered verdict). Verdict: none. Status: measured. Credit: Masters & Widmer 1995, AGU, cited. Used for research under the source's published terms; written confirmation is pending. Licence: the publisher's published terms.
2026-10-04 · reading · none The microwave sky's first three acoustic peaks, read by a fixed rule. Reading the published Planck 2018 temperature spectrum by a rule fixed in advance, the first three acoustic peaks sit at multipoles 223.357, 535.214 and 818.974. Limit: A fixed-rule reading of someone else's published spectrum: it tests nothing about our model, and the cosmology checks of our model are registered, not run. Record: fixtures/spectra/universe_2026_10_04.json. The record's word: basis: public_data (a reading; no registered verdict). Verdict: none. Status: measured. Credit: Based on observations obtained with Planck (http://www.esa.int/Planck), an ESA science mission with instruments and contributions directly funded by ESA Member States, NASA, and Canada. Planck Collaboration 2020, 'Planck 2018 results. VI. Cosmological parameters', A&A 641, A6. Used for research under the source's published terms; written confirmation is pending. Licence: the publisher's published terms.

Releases

2026-10-06 · release Six papers and eighteen pictures went out. The research page lists the six papers, each published in full with its figures, and the gallery shows the eighteen pictures, each drawn from a run the engine made. Limit: This line says what the release contained, counted from the record; it is not a measurement of how the pages performed. Counted from: deploy/papers_graph.json, deploy/gallery_curation.json.
2026-10-08 · release A seventh paper is published. The seventh paper, HOPE: recall in a library of Hilbert spaces, what it finds and where it stops, is published on agileintegrations.net. Limit: The paper registers no test yet, so it adds no registered prediction. Read it