Start here — what is on this page
Hypothesis machine — random recombination, scored for testability. A toy with a point.
Base rate of prior art — sixteen claims checked, sixteen already published. The ceiling, not the rate.
One failure, many names — 6,000 entries, 60 domains: the same error under different words.
Lab notebook — what was tried, dated, including what failed.
Run it in 90 minutes — the single session, timed minute by minute, before you commit to a day.
Start here
Research overview — what is published, and the rules it is held to.
Magnitude-scale mixing and the b-value — 78,493 real events; per-scale b spans 0.70–2.23 while the pooled value is stable.
Support the research — everything here is free; this is how it stays that way.
Corrections register — sixteen claims withdrawn or narrowed, with the mechanism of each error.
Test kit — seven runnable checks that audit a catalogue for those same defects.
Run the audit — paste your own numbers and get the checks back, in the browser.
Failure atlas — search 1,200 measurement failure modes by symptom rather than by discipline.
research_tools — 58 dependency-free helpers for numeric research data, documented from source.
Science & medicine · Open research notebook
Questions. Tests. Evidence.
Follow scientific and medical hypotheses from the question to the test and the reported result. Explore DeepSeek’s developing research, with data, methods and uncertainty kept together.
Audit log — external checks, kept current
4 November 2026 · running recordAn outside audit is running against this archive. Every check is logged in one place as it finishes, newest first, with what was tested, what came back, and whether anything here changed. Entries are appended; none is deleted.
Coverage so far: the storage premium does not survive a spectrum-preserving null; the hydrogel negative result holds and is classical; the #13 magnetic link budget is correct and its binding unknown is a cell-level growth-time constant, not geometry; H203 reproduces and needed error bars rather than correction; the #24 tactile budget is arithmetically exact but 90 % one unmeasured assumption; E4-H160’s √2 correction is confirmed and its open two-target question now has a number; the Blind Spots arithmetic yields one real error and one withdrawn accusation; and of the fifteen near-proof dossiers, thirteen are checkable, eleven are clean, and two corollaries are wrong.
Evidence level: independent computational re-analysis. Not an experiment, not a measurement, not peer review.
Storage premium tested against a full-spectrum null
4 November 2026 · independent re-analysisThe 10.7 % storage premium does not survive a null that preserves the full autocorrelation. The dossier’s own limits section says a phase-randomised surrogate would be a stronger test and had not been run; it has now been run. A surrogate that reproduces the exact amplitude distribution, the complete power spectrum and the seasonal cycle — and destroys only the ordering — leaves a median premium of +2.1 %, with 0 of 10 years below the 5th percentile and a sign test at p = 0.17.
Read plainly: the published number measures real weather against a counterfactual with no multi-day persistence. It is a property of the residual-load power spectrum, not of chronology beyond it. The dossier’s mechanism finding is confirmed and strengthened; the headline is not supported.
Two years still stand out, and they are not the two the dossier nominates. 2023 is the most ordinary year in the set (+0.4 %, p = 0.50); 2017 is the outlier (+21.6 %, p = 0.06), and 2018 sits in the opposite tail (−45.8 %, p = 0.03).
H0346 hydrogel: the negative result holds — 0 of 224 layouts beat the area-only control — and the gain law transfers across every sweep, recovering Λ ≈ 0.50(a/δ)0.84 against the dossier’s 0.45. The mechanism is the classical access resistance of a single aperture, which is why concentrating open area into patches cannot win. An earlier draft of this check claimed the law was in-sample only; that was the audit’s error, corrected on the page.
Read both checks, with the numbers and the caveats →
Evidence level: independent computational re-analysis on existing model and data. No experiment, no measurement, not peer review.
Identifiability, and the comparison that decides the answer
4 November 2026 · independent re-analysisAn external audit re-ran this archive’s code, re-derived its formulas, re-solved one of its partial differential equations and tested two of its dossiers. Five findings, all computational, no experiment performed. One confirms the archive’s own derivation, one corrects the audit’s first draft, and three change what the next experiment should be.
- H0199 — the layout bias is arithmetic, not geometric. When the field is constant on the boundary the error is exactly (mask mean ÷ boundary value) − 1, at any number of witnesses. Two numbers read off an existing thickness map decide whether an edge-only design is disqualified. A 4 × 5 × 7 factorial also shows the estimator is a bigger lever than the layout: a flexible predictor on a random layout is the worst design tested, at 47–74 %.
- I2 — the model checks out; the bias claim still needs an estimator. An independent finite-volume solution confirms the audit page’s centre-temperature identity to 4.2 × 10−5, and its published table to 10−10. An earlier draft of this section claimed a missing factor 1/a; that was an error in the audit’s own reconstruction and was withdrawn the same day. What the check adds is the size of the effect: the cross-sensitivity between the two fitted parameters crosses unity at a* = 0.259, so measure at a ≲ 0.05.
- N0751 — drift is the binding requirement, not noise. Gain matching needs only 6.7 % and readout noise never binds, but linear drift must stay below 0.7 % of the swelling gain over the record. The matched-time control tests the easy requirement; the bench needs a simultaneous rigid reference in the same optical path.
- H034 — yes, distinct models fit the same data and disagree about defects. Models matching all three elastic constants to 10−4 disagree on the defect energy by 3.7× (9× with bond bending) and on the elastic dipole by 4–11×. The dipole the hypothesis wants as a predictor is the worst-determined object in the set. Twelve phonon frequencies at 1 % collapse the ambiguity to 1.3× — and the first experiment should be a published elastic + phonon fit, not a DFT campaign.
- #15 — the pulse penalty is a duty tax, not a frequency window. At matched throughput the plating-limited current is exactly d·jDC at every frequency, because plating fires at the peak and the peak current is set by duty alone. The 1 mHz window in the earlier model comes from the recovery term, not from transport; whether any penalty exists is decided by the re-intercalation half-life, which one rest-recovery assay measures in a week.
Two of the tests found that the proposed experiment was the wrong experiment. Three of the four errors across the audit share one cause: the comparison sentence was never written — matched RMS versus matched throughput, F(a) versus F(a)/a, an interaction cutoff versus matched elastic constants.
Read the five findings, the tables and the falsifiers → · Download the code and outputs →
Evidence level: independent computational re-analysis and model audit. Not an experiment, not a measurement, not peer review. Original claims and their statuses remain unchanged on the pages that carry them; this adds a later check rather than replacing the record.
Nov 3rd Research and Hypotheses
3 November 2026 · working collection344 research pages, 20 CSV files, and the original code, saved results and input data. Browse the complete collection by title or series, with reading guides, later corrections and novelty screening close at hand.
Includes mathematical arguments, computational tests, public-data reanalyses and untested hypotheses. Generated leads are questions, not established discoveries. The scientific programs were not rerun for publication; novelty is unverified.
Read and search the Nov 3rd collection → · Download original sources and data →
Current findings & next tests
Status notes · updated 2 November 2026Follow what changed, what the evidence supports and what would settle the remaining question. Source-selected leads are distinct from independently checked results. The latest eight-direction package reports no new physical experiments.
Supplied model / analytical work
Energy-storage chronology
Current finding: Re-scoped twice. The reported effect is strongly dependent on which weather persistence the comparison preserves; under a full-spectrum surrogate (4 Nov 2026) the premium falls to +2.1 % and is not significant, so the number should be read as a property of the residual-load power spectrum rather than of chronology.
Evidence: Supplied analysis of RTE observations with ideal-store and stylised fleet models. The smaller system-objective ratio is dominated by unserved-energy penalties, not demonstrated capital savings.
Review status and next milestone
Source-reported computations and derivations; not independently reproduced during publication or this editorial review.
Next: Validate time handling and optimization independently; test whether exceptional-year residuals survive stronger persistence-preserving controls.
Novelty: Not established for the exact contribution in this status record.
External review: No outside specialist review is documented in this status record.
Supplied model / analytical work
H0199 · Reference-witness placement
Current finding: Saved numerical contradictions reconciled; random-field comparisons require a corrected rerun.
Evidence: Saved-output arithmetic and static code review clarify the finite comparisons. No new simulations or deposition measurements; random-field normalization remains unresolved.
Review status and next milestone
Source-reported computations and derivations; not independently reproduced during publication or this editorial review.
Next: Repair common-field normalization and check a separately versioned rerun before using random-field design rules; physical validation remains open.
Novelty: Not established for the exact contribution in this status record.
External review: No outside specialist review is documented in this status record.
Earlier editorial context → · Source report
Supplied model / analytical work
N0751 · Differential swelling sensor
Current finding: A well-specified instrument hypothesis with a strong matched-time control; predicted performance remains unmeasured.
Evidence: Analytical two-path, first-order response plus simulations with assumed diffusion and readout noise. No device measurements.
Review status and next milestone
Source-reported computations and derivations; not independently reproduced during publication or this editorial review.
Next: Measure each pathway’s gain, response time and noise; freeze parameters before unseen-waveform and matched-time-control comparisons.
Novelty: Not established for the exact contribution in this status record.
External review: No outside specialist review is documented in this status record.
Seven more developed investigations Thermal metrology, optical ranging, materials, magnetics, hydrogel, pulse charging and tactile calibration
Independent computational audit · 2 November 2026
I2 · Thermal-metrology model
Current finding: the stated heat-model solution survives independent checks; a universal apparent-conductivity bias and novelty claim are not established.
Evidence: separate derivation, finite-volume/transform checks and primary-literature comparison. No observations were analyzed or physical experiment performed.
Review status and next milestone
AI-assisted independent model check, not external peer review. Beam conventions, the inverse relation and source uncertainty response require correction. Broad variable-spot methods predate this work.
Next: obtain a compatible calibrated raw temperature/spot-size series for the frozen held-out protocol; no dataset was verified eligible.
Novelty: not established; substantial prior art identified.
Independent local model check
E4-H160 · Single-target delay precision
Current finding: A bounded mathematical correction with a separately implemented local statistical check; not evidence of a new physical resolution limit.
Evidence: Corrected complex-noise Fisher information and known-versus-nuisance phase models; independent seeded local high-SNR Monte Carlo. No physical data, global acquisition test or multi-target validation.
Review status and next milestone
Earlier publisher correction records an independently authored local simulation and arithmetic recalculation. It did not reproduce the supplied FFT estimator or physical experiment; no rerun in this editorial review.
Next: Keep corrected tables and assumptions adjacent to the original report; broader ranging claims require a model of multiple targets, nuisance parameters and acquisition failures.
Novelty: Not established for the exact contribution in this status record.
External review: No outside specialist review is documented in this status record.
Supplied model / analytical work
E4-H034 · Elastic dipoles and MLIP error
Current finding: The original missing-far-field-energy explanation is unsupported; reference-assisted error prediction remains an open lead.
Evidence: Supplied lattice-model comparisons, not DFT or MLIP validation on real defects. The required dipole reference is not a free runtime feature.
Review status and next milestone
Source-reported computations and derivations; not independently reproduced during publication or this editorial review.
Next: Repair the pilot split: defect-family/configuration groups can test one material; held-out-material claims require multiple materials.
Novelty: Not established for the exact contribution in this status record.
External review: No outside specialist review is documented in this status record.
Supplied model / analytical work
#13 · Magnetic battery warning
Current finding: Magnetic detectability is worth a controlled apparatus check; battery warning time is not established.
Evidence: Field calculations and idealized background rejection. Return geometry, growth rate, failure threshold and false-alarm behavior remain assumed or unmeasured.
Review status and next milestone
Source-reported computations and derivations; not independently reproduced during publication or this editorial review.
Next: A wire phantom may establish signal/background discrimination only; real fault-labelled battery measurements would be needed to assess warning time.
Novelty: Not established for the exact contribution in this status record.
External review: No outside specialist review is documented in this status record.
Supplied model / analytical work
H0346 · Perforated hydrogel skin
Current finding: A useful negative result within the stated diffusion model; retain it beside the original favorable claim.
Evidence: Supplied simulations do not support the proposed gain over the area-only control. Constant-diffusion and idealized geometry assumptions exclude some real-gel mechanisms.
Review status and next milestone
Source-reported computations and derivations; not independently reproduced during publication or this editorial review.
Next: State the conditional refutation prominently; an independently controlled diffusion measurement could determine whether model assumptions describe a real sample.
Novelty: Not established for the exact contribution in this status record.
External review: No outside specialist review is documented in this status record.
Supplied model / analytical work
#15 · Sodium-ion pulse charging
Current finding: The original universal penalty framing is replaced by a mechanism-dependent comparison.
Evidence: Conditional charge-transfer mathematics and an uncalibrated lumped-cell model. No cell measurements; the proposed frequency window is parameter-dependent.
Review status and next milestone
Source-reported computations and derivations; not independently reproduced during publication or this editorial review.
Next: Specify matched charge throughput and temperature controls before testing plating onset; do not present the model window as a measured cell property.
Novelty: Not established for the exact contribution in this status record.
External review: No outside specialist review is documented in this status record.
Supplied model / analytical work
#24 · Tactile-force calibration
Current finding: A useful uncertainty-budget proposal; precision and standards compliance remain unvalidated.
Evidence: Calculated budgets use assumed hysteresis, creep, geometry and material properties. No measured phantom or sensor calibration; standards were not fully assessed.
Review status and next milestone
Source-reported computations and derivations; not independently reproduced during publication or this editorial review.
Next: Measure one sensor’s history-dependent response and characterize the reference material before claiming an achieved uncertainty or certified standard.
Novelty: Not established for the exact contribution in this status record.
External review: No outside specialist review is documented in this status record.
All current assessments, source reports and editorial errata →
Weathering coupons · #16: retained as a proposed field experiment. No newer completed result was established in this review; read the original proposal.
Three levels, with the evidence stated separately
1 · Hypotheses
The full question pool, grouped and searchable. Ideas need no proof to be preserved.
2 · Developed leads
Source-selected proposals with supporting context and proposed tests. Selection depth does not establish correctness or novelty.
3 · Results & checks
Completed work labeled as mathematics, simulation, observation or experiment. A model result supports only its stated assumptions.
Lead and dossier selections are nested memberships, not additional hypotheses. Counts belong to their named release; a deduplicated project-wide total has not been established. Bibliographic context is not verified evidence for an exact effect.
Original wording, negative results, counterexamples and prior versions remain available. Archive inclusion does not certify truth, novelty or external peer review. Blogger is reserved for bounded results after adequate checks.
Discovery, investigation & validation
Explore broadly, develop selected questions deeply, and seek independent evidence. Our emerging focus is overlooked measurable effects: what a reference, instrument, model or time average can hide. The wider hypothesis archive remains open to other scientific directions.
Four research directionsMeasurement blind spots · novel sensing · materials and transport · time-dependent mechanisms
Measurement blind spots
Look for effects hidden by reference placement, inaccessible parameters or corrections that remove useful information. H0199 is a candidate for a controlled reference-layout comparison; inverse-problem proposals also belong here.
Sensing & instrumentation
Combine response times, geometries and sensor modalities to distinguish competing mechanisms. N0751 offers a specific analytical response and matched-time control; device performance remains unmeasured.
Materials, interfaces & transport
Compare defects, diffusion and boundary behavior with suitable reference measurements. H034 needs defensible reference data; the negative H0346 model finding remains part of the record.
Time, sequence & history
Test when persistence, ordering, frequency or loading history changes behavior despite matching static totals. Storage chronology, pulse charging and tactile calibration provide current examples.
What comes nextA proposed research route, not a claim that experiments or partnerships are underway
- Reliable baselines: H0199’s numerical reconciliation is published; its random-field implementation remains unresolved. Independently check selected storage and hydrogel calculations with explicit tolerances.
- Decisive test packages: prepare H0199 and N0751 with competing predictions, apparatus, controls, stop criteria and required laboratory capabilities. Establish reference-data feasibility before advancing H034.
- New observations: physical measurements require an appropriately equipped collaborator. No laboratory engagement, purchase or completed experiment is claimed by this plan.
- Report what survives: retain supported, contradicted and inconclusive outcomes, then assess the precise contribution against primary literature.
Idea generation continues alongside selective investigation and validation. Promotion follows checks of the particular result; archive inclusion does not depend on a favorable outcome.
How to read an evidence recordEvidence type, reproduction, external review and novelty are separate questions
- Hypothesis
- A proposed explanation or prediction; untested unless specific evidence is stated.
- Analytical result
- A derivation within explicit assumptions. Correct mathematics can describe an unsuitable physical model.
- Computational result
- A calculation or simulation with recorded inputs, code, controls and numerical limits.
- Measured evidence
- Identify observed inputs separately from a new experiment, including provenance, calibration and uncertainty. Using measured source data does not mean a proposed device was tested.
- Independent reproduction
- State who or what separate implementation checked which claim, with matching inputs and tolerances. An independent model check is not physical replication.
- External review
- Identify the outside specialist and review scope when documented. Publication, AI critique and code checking are not external peer review.
- Novelty
- Record a separate primary-literature comparison for the exact contribution. Unassessed, known overlap and unresolved originality are valid states; a mathematical proof alone does not establish novelty.
Keep the original hypothesis, its identifier, revisions and negative findings linked together. These fields describe different dimensions of evidence; they are not a ladder where one label replaces the others.
Latest research packages
Latest packages & correctionsSet 8, Discovery07, Measurement Blind Spots, developed directions and correction records.
I2 + H0199 audit kit
3 November 2026 · reproducibility scaffold, not an audit of actual results
The kit checks I2 numerical outputs and H0199 saved-versus-recomputed placement scores when the original equations, raw data, placements and fresh results are supplied. Its synthetic software tests do not validate Incoprea findings.
Read the audit-kit notice and direction → · Download the preserved kit →
Set 8 · Practical Inventions
2 November 2026 · source-preserving invention proposals
400 proposals across twenty capability domains and twenty physical approaches, with 100 screened entries, twenty editorial developments and three supplied engineering studies. Six developed concepts overlap the screened set. No prototype was built; costs and performance are estimates or model outputs, with report/output discrepancies noted.
Search all 400 proposals and expand the development plans → · Dew harvesting, compliant emitters and passive cooling → · All 22 originals, corrections and source methods →
Discovery07 · The Black Swan Grid
2 November 2026 · source-preserving exploratory collection
900 original hypotheses across 30 operators and 30 domains; 300 mechanically screened entries, 30 editorial refinements and three supplied model reports. Twenty-one refinements overlap the screened set; all selections retain their original identifiers. No experiment, observed-data analysis or independent scientific reproduction is reported.
Search the 900 ideas, expand refinements and read the supplied calculations → · Membership and result limitations → · All 26 original files, methods and corrections →
I2 independently checked · open measurement tests
2 November 2026 · mathematical/model audit and proposed protocols
The Gaussian heat-model solution agrees with independent numerical checks, but beam/readout conventions, an inversion and inference claims need correction. Primary literature already describes variable-spot methods. No compatible raw observation dataset was established eligible for the frozen test.
Findings, derivation, literature and reproducibility package → · H0199 and N0751 physical-test packages → · Invitation for equipment, access and data →
H0199 · saved-output correction
2 November 2026 · arithmetic reconciliation and static code inspection; no new simulations or physical experiment.
The saved radial-field count sweep gives 1.609537% RMSE for four grid witnesses and 0.773497% for 36 edge witnesses: the original cross-count comparison is reversed. The main linear-field comparison is not a tie, and a degree-1 fit reverses the edge/grid ranking. Random-field normalization also needs repair before those comparisons support design rules.
Read the separate correction, reconciled tables and scope → · Original report, preserved →
Measurement Blind Spots · new collection
Supplied 1 November 2026 · analytical / computational300 hypotheses, 30 nested refined leads and three supplied investigations across six scientific domains. Original identifiers, baseline assumptions, distinguishing observations and refutation conditions remain together. No experiment or direct published-data reanalysis was performed; the supplied calculations have not been independently reproduced during publication.
Explore the complete collection → · Search all 300 hypotheses · Read the 112-page PDF
Investigations, negative findings and original files
- I1 · Decay fitting: offset and finite-window effects, analytical identities and parameter-identifiability limits.
- I2 · Thermal metrology: a conditional boundary-conductance model predicting spot-size-dependent apparent conductivity.
- I3 · Impedance: a nonlinear model whose retained fundamental spectrum can match a Debye form. Read the correction history before quoting Lin–KK claims.
The source retains a narrowed OD600 stray-light hypothesis, corrected uncertainty interpretations and numerical fixes. Novelty scores and literature-access claims are supplied assessments, not independent clearance. Export gaps and class-label discrepancies are documented with the preserved originals.
30 developed leads → · Three supplied investigations → · Corrections and provenance → · All 27 original files →
New on the Research blog: A guide to the research page → · Three updates where the claim became narrower →
Eight directions · developed further
Latest supplied research · 1 November 2026Eight reports develop the earlier priorities through source-data reanalysis, conditional mathematics, simulation and experiment design. Negative findings and revised claims are retained. No new physical experiments were performed; the supplied calculations were not independently rerun during publication.
Explore all eight reports, figures, code and results →
Storage: show the sensitivity curve
The supplied report reproduces the earlier roughly 11% comparison, then reports 5.1% and 1.8% storage premiums under 3-day and 7-day persistence controls. A separate system-cost model reports a 1.3% effect. These are distinct comparisons, not interchangeable storage estimates.
H034: a mechanism correction
The supplied lattice study challenges the missing-far-field-energy interpretation. Its cutoff trend changes with parameterisation; real MLIP predictive value still requires a defensible data pilot.
Designs and negative results
Sensor, placement, magnetic-warning, hydrogel, sodium-pulse and tactile dossiers retain controls, failure conditions, modeled costs and next experiments. Estimated performance and budgets are not measured validation.
Open each direction and the evidence record
- Storage chronology · sensitivity to persistence controls →
- H034 · model mechanism challenged; DFT pilot proposed →
- N0751 · analytical sensor design and decisive control →
- H0199 · placement model and boundary failure case →
- #13 · magnetic warning budget and background rejection →
- H0346 · negative result across 208 modeled layouts →
- #15 · mean/peak pulse distinction under matched throughput →
- #24 · uncertainty chain and revised failure criterion →
Evidence levels · Supplied site review · All 139 original files
H034 & H160 · Corrections and independent checks
1 November 2026 · bounded publisher checks
H034: the prose energy table is fourfold too high; all 12 original CSV rows already match the formula. H160: the declared complex-noise convention requires twice the Fisher information, reducing the standard-deviation bound by √2. A separately written local Monte Carlo check is consistent with that correction; its scope is single-target precision.
Corrections, evidence and remaining questions
H160’s T2 prose precision table also has a factor-ten transcription error. The broad scanning/ranging tradeoff has verified prior art. Unknown target phase, two-target resolution, physical target structure and novelty remain separate questions. No real-data H034 pilot was performed.
Read the correction tables, independent code, trial results and primary-source comparison →
Research archive & reference library
Browse by collection, notebook, result package or research history. Nothing is removed when a claim changes; earlier text retains its dated context.
This searches the page’s findings and archive descriptions. Open a study collection to search its full hypothesis records.
Searchable study collectionsStudies 1–2, the 900s, Expeditions I–V, idea generators and the partial v9 snapshot.
Set 8 · 400 searchable invention proposals and development plans →
Discovery07 · 900 searchable hypotheses and source-selected development →
Research in three levels
New archive · Studies 1 & 2 · 31 October 2026Keep the whole question pool, show the most developed leads, and highlight results with their exact evidence level. These two supplied studies are organized collectively into five reader pages; every original file is retained.
4,200 proposals
Grouped, searchable and expandable. A raw idea does not need proof to be archived. Original hypotheses, tests and contrary outcomes stay together.
450 shortlisted leads
Includes 45 deeper dossiers. Source-selected feasibility and literature context, with access limits. These selections are not probabilities of truth.
5 conditional model results
Complete written derivations, reproduced code checks, assumptions and counterexamples. Physical hypotheses and novelty remain unverified.
Counts are nested: the leads, dossiers and model results are subsets of the 4,200 hypotheses. All 741 supplied files—including unselected proposals, replaced choices, blank templates and negative cases—are preserved in the original archives.
Open the complete compact archive → · Download originals and provenance →
Results on Blogger: Three model results from Study 1 → · Two model results from Study 2 →
Archive availability does not certify a claim. Evidence review controls highlights and Blogger publication, while the source record remains available. Reddit is reserved for unusually strong, appropriately supported results.
900 Studies: hypotheses, leads and focused dossiers
New archive · 1 November 2026The new Desktop 900s intake is preserved in three nested levels: the full hypothesis pool, refined evidence-linked leads, and focused dossiers or supplied result packages. Older publications remain untouched.
8,100 hypotheses
All 900-level version folders are grouped and retained. Raw ideas are archived without requiring proof.
810 refined leads
Editorial or evidence-linked selections with their source context and limits. These are not truth probabilities.
81 level-3 packages
Focused protocols, conditional analyses or supplied result packages. The supplied validation files report no physical measurements or completed proof of the underlying hypotheses.
Everything supplied is retained: 425 files, including alternate versions, validation records, search logs, source maps, code, blank observation tables and supporting notes.
Open the 900 Studies archive → · Search the Discovery Library & pipelines → · Download the complete frozen originals ZIP →
This is working research material, not peer review or clinical proof. Blogger remains reserved for strong reproducible results; this intake is therefore website-only for now.
Two new discovery expeditions
600 hypotheses · 60 nested refined dossiers · six supplied mathematical/computational result reports.
Search both 300-study expeditions →
Methods, corrections and original files
Each expedition preserves its 300 proposals, 30 refined dossiers, three supplied reports, corrections and refutations. Supplied results include partial conclusions and conditional model proofs; publication does not establish experimental validation.
Expedition III · The plausible impossible
300 hypotheses · 30 nested refined dossiers · three supplied result reports.
Corrections, refutations and original files
The supplied investigations report a corrected barrier-transmission law, a network-flow hypothesis refuted as stated, and a conditional pulse-coupling result. Complete original wording, code, limits and source files remain accessible.
Expedition IV · Convergent discovery
300 hypotheses · 30 nested refined dossiers · three supplied result reports.
Results, limitations and originals
Supplied mathematical and computational investigations include an inconclusive report. Complete source wording, corrections, code and limitations are preserved with 97 scientific source files.
New research · Versions 5, 8 and 9
Explore the three new archives
Version 5 · The hidden discovery laboratory →
300 hypotheses, 30 nested dossiers and three investigations in a supplied synthesis.
Version 8 · Ideas for new ideas →
175 generators, inventions, open questions and exploration items across five groups.
Version 9 · AI discovery methods →
Partial framing snapshot: three original files; study sections had not arrived at capture. All three archives preserve original files and source status.
Original notebooks & idea portfoliosVolumes I–XXXIII, Frontier Ideas and supplied proposal portfolios. Expand a volume group to browse.
Original theory notebooks Expand / collapse
Original theory notebooks
Updated 30 October 2026 · 47 retained science questions across 2 source volumes. Working hypotheses, not established findings or peer-reviewed conclusions. Original entry numbers are preserved.
From cosmology to origins
Questions spanning gravity, evolution, cognition, climate, physiology, materials, information and astrobiology. Each entry retains its proposed mechanism, predictions and refutation conditions.
Volume II · 23 science questions
Retained questions include Astronomy & Planetary Science, Chemistry & Reaction Networks, Mathematics & Computation, Engineering & Materials Failure, Psychology & Behaviour. The filtered index preserves original entry numbers, mechanisms, predictions and refutation conditions.
Original theory notebook · Volume III Expand / collapse
Original theory notebook · Volume III
Updated 30 October 2026 · 19 retained science questions across 1 source volume. Working hypotheses, not established findings or peer-reviewed conclusions. Original entry numbers are preserved.
Volume III · 19 science questions
Retained questions include Geology & Volcanology, Energy & Power Systems, Sleep & Chronobiology, Archaeology & Human Origins, Meteorology & Severe Storms. The filtered index preserves original entry numbers, mechanisms, predictions and refutation conditions.
Original theory notebook · Volume IV Expand / collapse
Original theory notebook · Volume IV
Updated 30 October 2026 · 26 retained science questions across 1 source volume. Working hypotheses, not established findings or peer-reviewed conclusions. Original entry numbers are preserved.
Twenty-six more questions
From physiology, vision and movement to ice, fungi, insects, cities, manufacturing, software security and radiation. Each entry presents a proposed explanation, three predictions and a result that would refute it.
Original theory notebook · Volume V Expand / collapse
Original theory notebook · Volume V
Updated 30 October 2026 · 18 retained science questions across 1 source volume. Working hypotheses, not established findings or peer-reviewed conclusions. Original entry numbers are preserved.
Volume V · 18 science questions
Retained questions include Reproductive Biology, Haematology & Coagulation, Aviation & Aerospace, Marine Biology & Fisheries, Hearing & Auditory Science. The filtered index preserves original entry numbers, mechanisms, predictions and refutation conditions.
The notebook grows · Volumes VI–XI Expand / collapse
The notebook grows · Volumes VI–XI
Updated 30 October 2026 · 281 retained science questions across 6 source volumes. Working hypotheses, not established findings or peer-reviewed conclusions. Original entry numbers are preserved.
Volume VI · 18 science questions
Retained questions include Radiology & Medical Imaging, Oncology, Public Health & Epidemiology, Tribology & Wear, Transport & Logistics. The filtered index preserves original entry numbers, mechanisms, predictions and refutation conditions.
Volume VII · Science questions
Skin barriers, oral biology, exercise physiology, food science and fermentation.
Volume VIII · 85 science questions
Retained questions include Engineering & Materials, Energy & Environment, Medicine & Health, Earth & Space, Mind & Behaviour. The filtered index preserves original entry numbers, mechanisms, predictions and refutation conditions.
Volume IX · 38 science questions
Retained questions include Life, Medicine & Mind, Society, Computation & Engineering, Physics, Cosmos & Matter, Earth, Climate & Energy. The filtered index preserves original entry numbers, mechanisms, predictions and refutation conditions.
Volume X · Science questions
An installment of 120 proposals across twelve domains, including acoustics, optics, robotics, water, materials, veterinary medicine and forests.
Volume XI · 8 science questions
Retained questions include Anesthesiology & Critical Care, History & Historical Dynamics, Rheumatology & Connective Tissue, Nephrology & Renal Physiology, Conservation & Heritage Science. The filtered index preserves original entry numbers, mechanisms, predictions and refutation conditions.
Further proposals · Volumes XII–XIV Expand / collapse
Further proposals · Volumes XII–XIV
Updated 30 October 2026 · 100 retained science questions across 3 source volumes. Working hypotheses, not established findings or peer-reviewed conclusions. Original entry numbers are preserved.
Volume XII · Science questions
Twenty-nine physical-science subfields, including plasma, fluids, soft matter, chemistry and the physical Earth.
Volume XIII · Science questions
Sixty-two proposals across 31 subfields, including acoustofluidics, optomechanics, spintronics, superconductivity and materials.
Volume XIV · 9 science questions
Retained questions include Consciousness & Attention, Transplant & Immune Tolerance, Regenerative Biology. The filtered index preserves original entry numbers, mechanisms, predictions and refutation conditions.
More questions · Volumes XV–XVIII Expand / collapse
More questions · Volumes XV–XVIII
Updated 30 October 2026 · 252 retained science questions across 4 source volumes. Working hypotheses, not established findings or peer-reviewed conclusions. Original entry numbers are preserved.
Volume XV · Science questions
Clinical and applied domains, including space medicine, toxicology, rehabilitation and trial methods.
Volume XVI · 92 science questions
Retained questions include Competing Rates, Geometry & Scale, Thresholds & Tipping, Networks & Connectivity, Selection & Bias. The filtered index preserves original entry numbers, mechanisms, predictions and refutation conditions.
Volume XVII · Science questions
Thirteen engineering disciplines, from coastal and earthquake engineering to railways, water, buildings and aerospace.
Volume XVIII · Science questions
Eleven habitats and built environments, from sea ice and forests to rivers, coasts and cities.
Timescales and trade-offs · Volumes XIX–XX Expand / collapse
Timescales and trade-offs · Volumes XIX–XX
Updated 30 October 2026 · 86 retained science questions across 2 source volumes. Working hypotheses, not established findings or peer-reviewed conclusions. Original entry numbers are preserved.
Volume XIX · 46 science questions
Retained questions include Ultrafast, Seconds to Minutes, Weeks to Months, Millennia, Micro to Milli. The filtered index preserves original entry numbers, mechanisms, predictions and refutation conditions.
Volume XX · 40 science questions
Retained questions include Efficiency vs Resilience, Sensitivity vs Specificity, Plasticity vs Canalisation, Exploration vs Exploitation, Speed vs Accuracy. The filtered index preserves original entry numbers, mechanisms, predictions and refutation conditions.
Dimensionless numbers · Volume XXI Expand / collapse
Dimensionless numbers · Volume XXI
Updated 30 October 2026 · 39 retained science questions across 1 source volume. Working hypotheses, not established findings or peer-reviewed conclusions. Original entry numbers are preserved.
Thirty-nine more questions
Three proposals for each of thirteen dimensionless-number families, including Reynolds, Peclet, Damkohler, Biot, Mach, Froude, Weber and Strouhal numbers. Each claim retains its mechanism, prediction and refutation.
Scales and mixed questions · Volumes XXII–XXIII Expand / collapse
Scales and mixed questions · Volumes XXII–XXIII
Updated 30 October 2026 · 128 retained science questions across 2 source volumes. Working hypotheses, not established findings or peer-reviewed conclusions. Original entry numbers are preserved.
Volume XXII · Science questions
Ninety-nine proposals organized into nine scale families, from molecular and nanoscale processes through cells, ecosystems, planets and the cosmos.
Volume XXIII · Science questions
Twenty-nine deliberately ungrouped proposals with twenty topic tags. The original mixed-topic presentation and each prediction and refutation are preserved.
Further proposals · Volumes XXIV–XXV Expand / collapse
Further proposals · Volumes XXIV–XXV
Updated 30 October 2026 · 218 retained science questions across 2 source volumes. Working hypotheses, not established findings or peer-reviewed conclusions. Original entry numbers are preserved.
Volume XXIV · 189 science questions
Retained questions include Volcanic arcs, Asphalt rutting, Balance, Beer foam, Blushing. The filtered index preserves original entry numbers, mechanisms, predictions and refutation conditions.
Volume XXV · Science questions
Twenty-nine ungrouped proposals with source-supplied instrument or dataset labels. Labels are presented as supplied; publication does not verify their evidence.
More proposals · Volumes XXVI–XXVIII Expand / collapse
More proposals · Volumes XXVI–XXVIII
Updated 30 October 2026 · 196 retained science questions across 3 source volumes. Working hypotheses, not established findings or peer-reviewed conclusions. Original entry numbers are preserved.
Volume XXVI · 105 science questions
Retained questions include Addiction cycle, Agricultural growing cycle, Infrastructure lifecycle, Battery charge cycle, Bone remodelling. The filtered index preserves original entry numbers, mechanisms, predictions and refutation conditions.
Volume XXVII · 82 science questions
Retained questions include Biology & physiology, Earth & climate, Measurement & modelling, Transport & flow, Engineering & infrastructure. The filtered index preserves original entry numbers, mechanisms, predictions and refutation conditions.
Volume XXVIII · Science questions
Nine proposals with seven supplied topic labels, spanning systems across scales, measurement, method, evolution and physiology.
Testing routes and rival explanations · Volumes XXIX–XXX Expand / collapse
Testing routes and rival explanations · Volumes XXIX–XXX
Updated 30 October 2026 · 103 retained science questions across 2 source volumes. Working hypotheses, not established findings or peer-reviewed conclusions. Original entry numbers are preserved.
Volume XXIX · 77 science questions
Retained questions include Field programme, New analysis, Existing data, New facility. The filtered index preserves original entry numbers, mechanisms, predictions and refutation conditions.
Volume XXX · 26 science questions
Retained questions include The rival is a different mechanism, The rival is the same mechanism at another scale, The rival is a confound, The rival says the effect is not there. The filtered index preserves original entry numbers, mechanisms, predictions and refutation conditions.
Assumptions and cases · Volume XXXI Expand / collapse
Assumptions and cases · Volume XXXI
Updated 30 October 2026 · 83 retained science questions across 1 source volume. Working hypotheses, not established findings or peer-reviewed conclusions. Original entry numbers are preserved.
Volume XXXI · 83 science questions
Retained questions include Averaging removes noise, Extinction is gradual, Selection optimises, Bigger animals are slower, Correlation implies causation. The filtered index preserves original entry numbers, mechanisms, predictions and refutation conditions.
Further proposals and measurements · Volumes XXXII–XXXIII Expand / collapse
Further proposals and measurements · Volumes XXXII–XXXIII
Updated 30 October 2026 · 76 retained science questions across 2 source volumes. Working hypotheses, not established findings or peer-reviewed conclusions. Original entry numbers are preserved.
Volume XXXII · 54 science questions
Retained questions include Dissolution, Refinement, Inversion, Reversal, Displacement. The filtered index preserves original entry numbers, mechanisms, predictions and refutation conditions.
Volume XXXIII · 22 science questions
Retained questions include Measurement & inference, Networks & contagion, Failure & fragility, Life & ecology. The filtered index preserves original entry numbers, mechanisms, predictions and refutation conditions.
Frontier Ideas
Added October 30, 2026 · An idea book of 24 proposals, not a research paper. Not peer reviewed.
Explore the ideas and their decisive tests
A new collection spanning AI, medicine, neuroscience, materials, energy, quantum technologies, space and robotics. Each entry connects its cited starting points to a proposed experiment, failure modes and a result that would count against the idea.
New working hypothesis portfolios · 31 October 2026 Expand / collapse
New working hypothesis portfolios · 31 October 2026
Working hypotheses · Not peer reviewedThree newly arrived science-question portfolios are available as supplied working material. They are proposals for tests, not established findings, proofs, or peer-reviewed conclusions. Blogger publication remains limited to reviewed results.
Unconventional hypotheses · 20 questions
Original report, Markdown, machine-readable hypotheses, and sources. The archive preserves the supplied wording and provenance.
Original hypotheses from unconventional sources · 30 questions
Supplied Markdown, manifest, source notes, build script and readable index. No claim of validation is made here.
Unconventional hypotheses · second arriving set
A second 30-question supplied set with CSV, Markdown, sources, stylesheet and readable index. It remains a working hypothesis archive.
Publisher note: these portfolios are archived for transparency and future testing. Archive availability is not validation, novelty, peer review or proof.
Tests, results & reproduction archivesDated supplied reports, checked mathematics and earlier result packages; review status stays with each record.
Testing the notebook
Independent calculations and reproducible tests, alongside the original proposals.
30 October 2026 · Mathematical and computational model check
Bursts, backlog and queue stability
An exact queue calculation and 960 seeded simulations examine one mechanism proposed in Volume IV. The result distinguishes temporary backlog from instability under fixed-capacity assumptions; it does not establish real-road behaviour.
Read the test, proof, data and code →
30 October 2026 · Nominal-profile model test
Sequence changes ideal storage accounting
At fixed peaks and variance, the original order required a median 10.7% more nominal energy capacity than within-month day-reorder references in the primary scenario. The result uses a fixed source-label model, not a physically complete UTC timeline or an estimate of France’s actual storage needs. All five scenarios, data, code and limits are available.
Science and engineering results — 30 October 2026 Expand / collapse
Science and engineering results — 30 October 2026
Download the science results and reproduction archive (ZIP). Extract it and open research-results/index.html for readable reports, per-test review status, original working files and the reviewed reproduction packages.
Supplied working reports; original verdicts not validated. This archive contains natural-science and concrete engineering studies only. Original verdicts in working reports are not passed proofs. Held claims remain held; unreviewed tests and later revisions are identified. Country, demographic, economic, banking, polling, social, bibliometric and aggregate human population studies are excluded.
28 science collection/version packages; 108 completed test entries; 16 reviewed reproduction packages. Raw provider/full-text caches, credentials and runtime/scratch files are excluded.
Reviewed science and engineering findings
- Testing ensemble errors: a correlation floor, not a universal cutoff — Blogger published
- Testing cache size: the workload ratio matters, but access order matters too — Blogger published
- Testing service reliability: required dependencies matter more than graph depth alone — Blogger published
- When averaging becomes worse: an exact test of measurement drift — Blogger published
- Compact buildings lose less heat—until the envelope changes — Blogger published
- Quantum yield is a branching race, but one lifetime ratio is insufficient — Blogger published
- Shared failures cap redundancy, while every extra replica still helps — Blogger published
- A lower average latency can hide more missed deadlines — Blogger published
- Correcting a cache calculation: valid hit rates and a narrower ratio claim — Blogger published
- Drier modeled soil coincides with more extreme heat days across 50 city locations — Blogger published
- A supplied Poisson calculation: equal concentration does not imply equal relative noise — Blogger published
- A supplied Fisher–KPP check: wave speed depends on a product, not a ratio alone — Blogger published
- A supplied cluster count: corner and edge atoms occupy a larger fraction of smaller shells — Blogger published
- A supplied Euler calculation: an ideal column’s critical stress crosses a chosen reference — Blogger published
- A supplied Allee model: a threshold exists only within the bistable regime — Blogger published
- When the order of a sum first matters — Blogger published
Archived science working collections
| Collection | Completed tests | Review status |
|---|---|---|
| Incoprea Hypothesis Tests – Empirical Batch – 2026-10-30 — science-only subset | 2 | Needs review: 2 |
| Incoprea Hypothesis Tests III – 2026-10-30 — science-only subset | 10 | Held for correction: 5 · Blogger published: 5 |
| Incoprea Hypothesis Tests V – 2026-10-30 — science-only subset | 5 | Needs review: 5 |
| Incoprea Hypothesis Tests VI – 2026-10-30 — science-only subset | 1 | Needs review: 1 |
| Incoprea Hypothesis Tests VII – 2026-10-30 — science-only subset | 7 | Needs review: 7 |
| Incoprea Probable Tests – 2026-10-30 — science-only subset | 7 | Held for correction: 4 · Blogger published: 3 |
| Incoprea Research – Probable Hypotheses Tested – 2026-10-30 — science-only subset | 3 | Held for correction: 2 · Blogger published: 1 |
| Incoprea Tested – 2026-10-30 — science-only subset | 7 | Needs review: 7 |
| incoprea-hypothesis-tests-iii-20261030 — science-only subset | 3 | Needs review: 3 |
| incoprea-v2-20261030 — science-only subset | 3 | Needs review: 3 |
| incoprea-vol2-20261030 — science-only subset | 9 | Needs review: 9 |
| incoprea-wave2-2026-10-30 — science-only subset | 2 | Needs review: 2 |
| seven-tests-source-freeze — science-only subset | 4 | Blogger published: 1 · Held for correction: 3 |
| seven-tests-source-freeze — later revision 1 — science-only subset | 3 | Needs review: 3 |
| seven-tests-source-freeze — later revision 2 — science-only subset | 3 | Needs review: 3 |
| supplied-hypothesis-tests-20261030-source-freeze — science-only subset | 1 | Held for correction: 1 |
| supplied-independent-tests-20261030-source-freeze — science-only subset | 1 | Held for correction: 1 |
| supplied-tests-II-20261030-source-freeze — science-only subset | 1 | Held for correction: 1 |
| Incoprea Hypothesis Tests – Computational Mechanisms – 2026-10-30 — science-only subset | 5 | Needs review: 5 |
| Incoprea Hypothesis Tests III Empirical – 2026-10-30 — science-only subset | 4 | Needs review: 4 |
| Incoprea Hypothesis Tests IV – 2026-10-30 — science-only subset | 3 | Needs review: 3 |
| hypothesis-tests-iv-20261030 — science-only subset | 5 | Needs review: 5 |
| incoprea-audit-b-20261030 — science-only subset | 1 | Needs review: 1 |
| incoprea-hypotheses-r2 — science-only subset | 3 | Needs review: 3 |
| incoprea-hypothesis-tests-2026-10-30 — science-only subset | 4 | Needs review: 4 |
| incoprea-round3-20261030 — science-only subset | 3 | Needs review: 3 |
| incoprea-round4 — science-only subset | 2 | Needs review: 2 |
| incoprea-tested-20261030 — science-only subset | 6 | Needs review: 6 |
Detailed per-test tables, source filenames and hashes are inside the ZIP. Further science results will be added in later releases. Return to the research hypothesis library.
Science and engineering supplement — 30 October 2026 Expand / collapse
Science and engineering supplement — 30 October 2026
Download the science results and reproduction archive (ZIP). Extract it and open research-results/index.html for readable reports, per-test review status, original working files and the reviewed reproduction packages.
Supplied working reports; original verdicts not validated. This archive contains natural-science and concrete engineering studies only. Original verdicts in working reports are not passed proofs. Held claims remain held; unreviewed tests and later revisions are identified. Country, demographic, economic, banking, polling, social, bibliometric and aggregate human population studies are excluded.
3 science collection/version packages; 10 completed test entries; 5 reviewed reproduction packages. Raw provider/full-text caches, credentials and runtime/scratch files are excluded.
Reviewed science and engineering findings
- A window below the bandwidth-delay product limits throughput smoothly — Blogger published
- A unit correction changes a probe-length calculation, but does not prove long probes fail — Blogger published
- A communication energy floor without a finite bandwidth optimum — Blogger published
- More water recovery costs more work in an ideal salt solution — Blogger published
- An intermediate band gap wins in the supplied ideal solar-cell model — Blogger published
Archived science working collections
| Collection | Completed tests | Review status |
|---|---|---|
| incoprea-batch3-20261030 — science-only subset | 1 | Needs review: 1 |
| incoprea-tests-2 — science-only subset | 4 | Needs review: 4 |
| review-tests — science-only subset | 5 | Needs review: 5 |
Detailed per-test tables, source filenames and hashes are inside the ZIP. Further science results will be added in later releases. Return to the research hypothesis library.
New supplied science results — 30 October 2026 Expand / collapse
New supplied science results — 30 October 2026
Download the science results and reproduction archive (ZIP). Extract it and open research-results/index.html for readable reports, per-test review status, original working files and the reviewed reproduction packages.
Supplied working reports; original verdicts not validated. This archive contains natural-science and concrete engineering studies only. Original verdicts in working reports are not passed proofs. Held claims remain held; unreviewed tests and later revisions are identified. Country, demographic, economic, banking, polling, social, bibliometric and aggregate human population studies are excluded.
2 science collection/version packages; 8 completed test entries; 0 reviewed reproduction packages. Raw provider/full-text caches, credentials and runtime/scratch files are excluded.
Review status
All eight newly supplied reports are Needs review; original verdicts are unvalidated, and no new proof is claimed here.
Archived science working collections
| Collection | Completed tests | Review status |
|---|---|---|
| Incoprea Hypothesis Tests – Eight Data & Model Tests – 2026-10-30 — science-only subset | 7 | Needs review: 7 |
| Incoprea Hypothesis Tests – ML & Simulation – 2026-10-30 — science-only subset | 1 | Needs review: 1 |
Detailed per-test tables, source filenames and hashes are inside the ZIP. Further science results will be added in later releases. Return to the research hypothesis library.
New supplied science results — 30 October 2026 Expand / collapse
New supplied science results — 30 October 2026
Download the science results and reproduction archive (ZIP). Extract it and open research-results/index.html for readable reports, per-test review status, original working files and the reviewed reproduction packages.
Supplied working reports; original verdicts not validated. This archive contains natural-science and concrete engineering studies only. Original verdicts in working reports are not passed proofs. Held claims remain held; unreviewed tests and later revisions are identified. Country, demographic, economic, banking, polling, social, bibliometric and aggregate human population studies are excluded.
8 science collection/version packages; 42 completed test entries; 0 reviewed reproduction packages. Raw provider/full-text caches, credentials and runtime/scratch files are excluded.
Review status
All 42 newly supplied reports are Needs review; original verdicts are unvalidated, and no new proof is claimed here.
Archived science working collections
| Collection | Completed tests | Review status |
|---|---|---|
| Incoprea Research – Hypothesis Tests V – 2026-10-30 — science-only subset | 9 | Needs review: 9 |
| Incoprea Research – Probable Hypotheses Tested II – 2026-10-30 — science-only subset | 14 | Needs review: 14 |
| hypothesis-tests-iv-20261030 — science-only subset | 5 | Needs review: 5 |
| incoprea-audit-b-20261030 — science-only subset | 1 | Needs review: 1 |
| incoprea-hypothesis-tests-2026-10-30 — science-only subset | 1 | Needs review: 1 |
| incoprea-round4 — science-only subset | 7 | Needs review: 7 |
| review-tests — science-only subset | 1 | Needs review: 1 |
| incoprea-tested-20261030 — science-only subset | 4 | Needs review: 4 |
Detailed per-test tables, source filenames and hashes are inside the ZIP. Further science results will be added in later releases. Return to the research hypothesis library.
Additional science working reports Expand / collapse
Additional science working reports
Needs review · 30 October 2026. Four newly completed result packages and one later supplied revision are now available. These source reports are awaiting independent consistency review; their original verdicts are unvalidated.
Download the five-report science supplement (ZIP) → Extract it and open research-results/index.html to read the reports, review boundaries, source code and input-availability notes.
- Antecedent rainfall and heat extremes · New completed result
- Bird mass and climate associations · Later supplied revision
- Solar-cell band-gap calculation · New completed result
- Communication energy and bandwidth · New completed result
- Gait cost and Froude scaling · New completed result
The archive preserves supplied code, result records and provenance. Upstream provider data and restricted input tables are omitted where noted; availability limits are stated with each report.
Ten hypotheses put to the test · 31 October 2026
Supplied working report · Not peer reviewedA completed, reproducible package checks ten mechanism-level science hypotheses using published measurements, a public weather archive, arithmetic, and conditional models. The supplied report records one model-supported result, four partial results and five not-supported results. These are working results, not endorsed proofs or peer-reviewed conclusions.
Download the complete ten-test archive →
The ZIP preserves the original claims, methods, code, data, figures, source extracts and machine-readable results. Computational checks are labeled as model or arithmetic checks; they are not experimental proof of nature.
Scope note: This package contains physical science, chemistry, biological mechanisms, natural-science observations and engineering mechanisms only. It is supplied working material and remains under review.
Three exact mechanism checks · 31 October 2026
Publisher-reviewed working proofs · Not peer reviewedThree supplied reports independently check questions already highlighted on this page. The results are exact mathematics or conditional model checks, not new experiments and not claims about nature beyond their assumptions.
Download the complete proof archive →
Q16: the Shannon floor is real, but the claimed finite optimum is wrong. Q10: Fisher–KPP invasion speed follows 2√(rD), not r/D. Q17: recovery work rises monotonically, while the claimed magnitude is wrong.
Publisher review checked the supplied scripts, tables and stated limits. The package is not externally peer reviewed.
New mechanism-level result packages Expand / collapse
SUPPLIED SCIENCE RESULTS · 31 OCTOBER 2026
New mechanism-level result packages
Completed supplied packages archived from the Desktop intake. These are publisher-screened working reports and exact/conditional model checks, not peer-reviewed claims or independent experimental proof. Original code, data, results, provenance and limitations are preserved in each archive.
- Three electrostatic hypotheses — tested models · Blogger findings index
- Three physical anomalies — audited working report
- Three exact mechanism proofs — conditional mathematics
- Ten tested hypotheses — supplied working results
- Three provable computational models — conditional mathematics · computational check only
- Three tested natural-science reports — supplied working archive · unreviewed
Publisher review checks scope, definitions, units, code/results consistency and limitations. Availability here is not validation, novelty, peer review or proof of nature. Research home.
Investigation history & earlier plansOlder storage summaries, original priorities, follow-ups and historical dashboards. Later findings appear above.
Historical record: this preserves an earlier assessment or plan. For the latest editorial position and later corrections, see current findings above. Quantities below retain their original comparison and review scope.
Featured investigations
Four projects to follow01 / Computational re-analysis · highest priority
Energy storage chronology →
The earlier annual comparison is followed by a new supplied persistence-control and system-model study. Its reported storage premium changes with the null; the older roughly 11% figure should be read alongside that sensitivity analysis.
Latest supplied storage re-analysis and null-sensitivity curve →
02 / Proposed field experiment
Weathering coupons · #16 →
Controlled buried-mineral samples, quantitative X-ray diffraction and comparison against independent carbon-removal measurements.
Next milestone: a small controlled pilot with an XRD-capable partner.
03 / Proposed measurement benchmark
Tactile force standard · #24 →
Calibrated sensors, reference forces and tissue-like materials for comparing robotic touch measurements.
Next milestone: compare existing calibration methods and test cross-sensor agreement under realistic materials.
04 / Proposed battery experiment
Battery dendrite warning · #13 →
Test whether magnetic signatures identify emerging lithium-metal battery failure before conventional measurements.
Next milestone: controlled measurements with a suitable battery and magnetometry laboratory.
Historical record: this preserves an earlier assessment or plan. For the latest editorial position and later corrections, see current findings above. Quantities below retain their original comparison and review scope.
New follow-ups · H034 and H160
Supplied working reports · 1 November 2026H034 · Empirical question still open →
A new derivation and synthetic evaluation protocol are available. No real defect dataset was assembled, so predictive value for real materials models remains inconclusive.
H160 · Conditional analysis and unresolved check →
The supplied report compares idealised point and extended targets. Its Monte Carlo check has an unresolved RMSE/CRB inconsistency; estimator attainment and novelty are unverified.
Read reports, corrections, code and original results
29 source files retain the original claims, preregistration, methods, limitations, CSV tables, logs and code. All 26 supplied checksum entries match. Research code was not executed during publication.
Historical record: this preserves an earlier assessment or plan. For the latest editorial position and later corrections, see current findings above. Quantities below retain their original comparison and review scope.
Original investigation plan · Expedition IV
Plan recorded before the follow-up reportsA supplied ChatGPT review proposes the following development priorities. Its literature comparisons and technical objections remain provisional until checked against the original reports and primary publications. This original plan is retained for chronology. Follow-up reports and the subsequent correction note appear above; remaining tasks include a real-data H034 pilot and broader H160 validation.
E4-H034 · Materials-model error prediction
Test whether an elastic-dipole descriptor adds predictive value for learned-potential errors. First establish compatible reference data; then compare with simple baselines and uncertainty estimators on held-out materials or defect families.
E4-H160 · Optical ranging tradeoff
Re-derive the exact array geometry and angular factors. Simulate the emitted field, target return and range reconstruction independently, and determine whether processing changes remove the proposed limit.
E4-H006 · Pinning and damping
Verify the closest literature and re-derive the momentum-dependent coupling and threshold behaviour before proposing a material-specific experiment.
Other candidates and correction tasks
E4-H209 · Molecular motor slip: identify a measurable diagnostic beyond established stepping models. E4-H154 remains a possible signal-processing follow-up.
C1: check prior-art overlap and identifiability of measured spectra separately from matrix rank. C3 retains its supplied inconclusive status.
The supplied review flags E4-H028 (oscillatory-tail convergence), E4-H135 (occupancy versus continuous-contact survival), E4-H108 (gravitational coefficient), and E4-H076 (charge magnitude) for source-level checking. These are reported review concerns, not newly verified corrections.
Each completed follow-up should retain original claims, corrections, negative outcomes, data provenance, code, controls and a dated account of what was independently reproduced. Novelty requires specific comparison with verified prior work.
Read the complete Expedition IV archive and supplied result reports →
Historical record: this preserves an earlier assessment or plan. For the latest editorial position and later corrections, see current findings above. Quantities below retain their original comparison and review scope.
Priority result and next tests
Focused upgrade path · supplied working resultSequence changes ideal storage accounting
Real RTE data (175,344 rows, 2015–2024) support a descriptive result: at fixed peaks and variance, sequence changes nominal storage capacity in all ten year profiles, with a median 10.7% difference in the primary scenario. The result uses nominal half-hour labels and a hypothetical 70/30 mix; it is not an adequacy forecast.
Upgrade path: rebuild the physical UTC timeline, justify the future mix, and compare the sequence effect against an actual adequacy model.
Weathering coupons · #16
Low-cost field coupons with quantitative XRD; focus on site-to-site variance and a decisive kill criterion.
NVD dendrite warning · #13
Classify the pre-collapse signature and measure lead time before voltage failure.
Pulse-aging sodium-ion · #15
Cycle cells under grid-like pulses rather than smooth laboratory current.
Tactile force standard · #24
A low-cost measurement standard with a direct calibration test.
Dual-sorbent water + carbon · #14
A practical paired-material test with a measurable separation trade-off.
The 33 theory volumes and larger question portfolios remain preserved below as frozen working archives. New supplied proposals may enter a clearly labeled archive. Evidence review controls promotion to highlights and Blogger; it does not remove raw ideas from the archive.
Historical record: this preserves an earlier assessment or plan. For the latest editorial position and later corrections, see current findings above. Quantities below retain their original comparison and review scope.
Top 10 investigations
Prioritized working resultsRanked by evidence quality, reproducibility, practical usefulness, originality and remaining uncertainty. These are working investigations, not peer-reviewed discoveries.
Sequence changes ideal storage accounting
Strongest current result: ten annual profiles agree in the primary scenario. Next: physical UTC timeline, defensible mix and adequacy-model comparison.
Quantum yield pathways
Branch-resolved rates matter; a single lifetime ratio is insufficient.
Averaging and drift
Independent noise falls with averaging while random-walk drift can accumulate against a fixed target.
Compact envelopes and heat loss
Geometry helps in the checked model, but windows, bridges and leakage can change the ranking.
Cell volume and signalling noise
At fixed concentration, smaller volumes imply larger relative Poisson fluctuations.
Fisher–KPP invasion speed
The checked speed scales as 2√(rD), not a simple ratio; this is a model result.
Particle shape and active-site fractions
Geometry changes corner and edge fractions; activity still requires controlled specimens.
Water recovery and ideal separation work
Higher recovery raises reversible work per product volume in the ideal-solution calculation.
Transport windows and throughput
A finite in-flight window limits ideal throughput through the window/RTT relation.
Cache workload ratio
Cache-to-working-set ratio matters, while reuse and popularity shape the curve.
Independent external review is not claimed. The next milestone is a reproducible challenge or replication, not more hypothesis volume.
Historical record: this preserves an earlier assessment or plan. For the latest editorial position and later corrections, see current findings above. Quantities below retain their original comparison and review scope.
Editorial triage: what is research-ready?
Updated 31 October 2026One flagship result, a methods notebook, and future experiments
The storage chronology re-analysis is the current flagship: a real-data engineering investigation with a concrete validation path. It should not be read as a national storage forecast until the physical UTC timeline and an adequacy model are completed.
The 20-question collection remains available, but textbook checks such as Fisher–KPP speed, Poisson noise, Euler buckling, Shannon bounds, Allee thresholds and cache ratios are now classified as a methods notebook: useful exact mathematics and error correction, not demonstrated scientific discoveries.
Research status: 108 reported tests, 16 reviewed reproductions, 0 claimed external peer reviews. Internal DeepSeek/Codex review is not independent validation.
Supplied evaluation note: A DeepSeek evaluation reports an independent recomputation of the storage statistic (median ratio about 1.111; 10/10 year profiles above its reorder null) and highlights convention dependence: approximately +10.9% for annual closure at β=1, +5.6% for decade closure at β=1, +2.7% for decade closure at β=0.5, and +0.7% for a fixed 4-hour battery/unserved-energy specification. This is a supplied editorial assessment, not external peer review or a new experiment; the figures remain tied to the stated model, data and assumptions.
Next validation: finish the storage UTC timeline, compare against a defensible adequacy model, place the weathering-coupon proposal before an XRD-capable group, and send the dendrite-warning proposal to a suitable magnetometry lab. No new hypothesis volume is the priority.
Historical record: this preserves an earlier assessment or plan. For the latest editorial position and later corrections, see current findings above. Quantities below retain their original comparison and review scope.
Research status dashboard
Publisher-maintained counts108
Completed entries currently listed in the science archive.
16
Packages with a documented reproduction review.
24
Proposals with a decisive test or kill criterion.
33
Frozen working archives, not established findings.
Storage chronology
Needs UTC validation and an adequacy-model comparison.
0 claimed
Internal publisher review is not independent validation.
Evidence labels used across the page: Proposal · Exact mathematics · Conditional model · Observational association · Experimental result · Independent review.
Historical record: this preserves an earlier assessment or plan. For the latest editorial position and later corrections, see current findings above. Quantities below retain their original comparison and review scope.
Storage chronology re-analysis · version 2
Held for major revision · working reportEvidence status: The annual within-month day-order measurement is reproducible in an independent reimplementation: median ratio 1.109964, 95% bootstrap interval 1.080455–1.168296, 10/10 years above the reorder-null median, exact sign-test p=0.001953. Peak and variance are invariant under the whole-day permutation by construction.
What is not endorsed: The decade/closure interpretation is held. The adversarial referee found that the pooled decade row is ordinary under its null (73rd percentile), the decade implementation balances each year separately, and the independent implementation does not reproduce the decade result. This package is therefore a supplied working report, not a validated national storage forecast or experimental proof.
Public-scope archive: Country-comparison datasets and country-specific raw extracts are excluded from this release and preserved privately. The science-only archive was frozen from the Desktop and external-volume aliases; SHA-256 a7d21fd684c76a1297d9eb80b5620cda22c96d0039181cfdaab508993fb1db24. The full source package remains available for further review through the local intake record.
Open the canonical storage test page → · Return to the Research archive →
Historical record: this preserves an earlier assessment or plan. For the latest editorial position and later corrections, see current findings above. Quantities below retain their original comparison and review scope.
Plain-English summaries
Read before the technical reportsThese summaries describe what the current records do and do not establish. Exact and model refer to stated assumptions; they are not experimental proof of nature.
Storage chronology
What did we discover? In the checked annual model, changing the order of otherwise identical half-hour demand and renewable values changed nominal storage capacity; the median difference was about 11% across ten year profiles.
Why might it matter? Storage sizing can depend on chronology, not only peak and variance, so a reported number needs its time window, balance rule and residual definition.
What remains unproven? This is not an adequacy forecast or physical proof. The closure interpretation is held for revision and needs a defensible UTC timeline, an adequacy model and external reproduction.
Methods notebook
What did we discover? The checked mathematics and conditional models catch arithmetic or definition errors and clarify which conclusions follow from the assumptions.
Why might it matter? Keeping code, inputs, units and limits beside each report makes the work easier to audit and correct.
What remains unproven? An exact calculation or conditional model does not show that nature, a device or a population behaves that way without measurements.
Priority experiments
What did we discover? Weathering coupons, dendrite-warning sensing, pulsed sodium-ion aging, tactile-force standards and paired sorbents define concrete tests with measurable outcomes.
Why might it matter? They turn the strongest proposals into low-cost, falsifiable next steps that another laboratory could run.
What remains unproven? These are proposals, not reported measurements. No new physical result is claimed until an independent test produces and documents the data.
Topic routes & Research blogEarlier subject guides, selected questions and existing Blogger findings.
The hypothesis notebook
Developing research · Not peer reviewedExplore the tests and reported findings
A growing record of scientific and medical questions tested by DeepSeek. Each entry keeps the question, method, result and limits visible, alongside the original files.
What was tested?
The hypothesis, dataset and comparison used in each test.
What did it show?
Source-reported support, results without support and unresolved questions, with the relevant numbers.
What can we inspect?
Methods, downloadable results, code and limitations. Exploratory associations remain preliminary.
Research and testing: DeepSeek. Organization and web publishing: Codex. Published entries are working reports; publication here does not establish scientific truth or clinical validity.
Latest science publications
Checked calculations and their limits; working research, not external peer review.
A communication energy floor without a finite bandwidth optimum
A normalized signal-energy floor, with no finite bandwidth optimum in the ideal channel. Assumptions and reproducible code remain with the result.
More water recovery costs more work in an ideal salt solution
Recovery increases reversible work in a specified ideal salt-solution model. Assumptions and reproducible code remain with the result.
An intermediate band gap wins in the supplied ideal solar-cell model
An intermediate band-gap maximum in the supplied radiative blackbody calculation. Assumptions and reproducible code remain with the result.
The working library
Direct routes into physical, biological and engineering questions, with source status and limitations kept visible.
Materials and flight
Earthquake records
Frequency and magnitude by region
Day and night catalogue comparison
Follow the evidence
From the Research blog
Quantum yield is a branching race, but one lifetime ratio is insufficient
Competing decay pathways, with the assumptions behind a lifetime-ratio shortcut made explicit.
When averaging becomes worse: an exact test of measurement drift
Averaging and drift checked against a defined measurement target.
A supplied Poisson calculation: equal concentration does not imply equal relative noise
A molecule-count calculation that separates concentration from relative noise.
Materials science
A lattice on paper
A supplied materials simulation. No specimen has been printed or physically tested.
Flight and Earth-science questions
Insect-scale flight calculations
A model calculation, with its assumptions and proposed tests.
Earthquake catalogues
Astronomical data
An exoplanet radius check
A working attempt to recover a known astronomical feature, with source corrections and limitations visible. It remains a preliminary analysis.
20 promising science questions Expand / collapse
20 promising science questions
Mechanisms · Tests · Next questionsSelected for clear mechanisms and checkable next steps across physics, chemistry, biology and engineering. These are questions worth pursuing, not twenty proven hypotheses. “Exact” means mathematics under stated assumptions; models and observations have their own limits.
Which decay pathways determine quantum yield?
Competing radiative and nonradiative rates determine yield in the checked rate model, making pathway-specific lifetimes worth separating.
Next test or remaining limit
Measure branch-resolved decay rates in a multiexponential emitter; a single mean lifetime ratio does not identify the yield.
When does more averaging increase measurement error?
Averaging suppresses independent noise but accumulates random-walk drift when estimating a fixed initial value.
Next test or remaining limit
Test a long measurement series against an independent reference and specify whether the target is the initial value or the current value.
How much can a compact envelope reduce heat loss?
Less exposed area reduces steady conductive loss in the checked envelope model, while surface-specific insulation can change the ranking.
Next test or remaining limit
Compare measured heat loss for matched floor area and insulation while accounting for windows, thermal bridges and air leakage.
When do shared errors erase ensemble gains?
Shared failures can defeat error averaging, but pairwise correlation alone does not specify every voting outcome.
Next test or remaining limit
Measure the full joint error pattern of an ensemble on held-out cases, including shared and higher-order failures.
Which workload properties control cache hits?
Cache size relative to working set matters in the supplied traces, while popularity and reuse patterns change the hit curve.
Next test or remaining limit
Compare real reuse traces at the same cache-to-working-set ratio to test how much locality changes the predicted hit rate.
Which required dependencies limit service reliability?
Reliability compounds across independently required services, so graph depth alone does not determine success.
Next test or remaining limit
Inject controlled failures into systems matched on required-service count to isolate topology and correlated outages.
Where does redundancy meet a shared-failure ceiling?
Replicas reduce independent failures but cannot remove an assumed common failure event, even though finite replica gains remain positive.
Next test or remaining limit
Estimate the common-event probability from joint replica failures rather than treating it as a chosen model parameter.
Can a faster average hide more missed deadlines?
The checked latency distributions show that average speed and the probability of exceeding a deadline can move in different directions.
Next test or remaining limit
Measure deadline exceedances in actual latency traces instead of assuming a lognormal tail.
How does cell volume change signalling noise?
At fixed concentration, a smaller volume contains fewer molecules and has larger relative fluctuations in the Poisson model.
Next test or remaining limit
Measure molecule-count fluctuations across cell volumes while checking whether the Poisson assumption fits the observed variance.
Which growth and dispersal combination sets invasion speed?
The Fisher–KPP calculation links front speed to the square root of growth times dispersal, challenging a simple ratio claim.
Next test or remaining limit
Measure spread in a controlled spatial population while estimating growth and dispersal independently.
How does particle shape change the fraction of active sites?
Geometric cluster counts change the relative abundance of corners, edges and terraces as particles become smaller.
Next test or remaining limit
Measure catalytic turnover for particles with controlled shape and composition; site counts alone do not establish activity.
When should a slender limb buckle before it bends?
Euler scaling makes long slender supports susceptible to buckling, with strength depending on geometry and end constraints.
Next test or remaining limit
Compare measured failure loads with the model using the specimen’s actual cross-section, material stiffness and boundary conditions.
When does an ecological population need a minimum starting density?
The checked Allee model has an unstable density threshold because low-density reproduction cannot offset losses.
Next test or remaining limit
Test replicate biological populations started at different densities and estimate the birth and loss curves independently.
When does summation order change a numerical answer?
Finite precision makes rounding depend on the accumulation path in the reproduced dyadic-sum calculation.
Next test or remaining limit
Check other input families and hardware reductions against exact arithmetic before proposing a universal depth threshold.
How strongly does dry soil accompany heat extremes?
The supplied reanalysis association is consistent with reduced evaporative cooling when soil dries, without isolating that causal mechanism.
Next test or remaining limit
Validate the relationship in independent observations and years, separating soil moisture from the weather conditions that influence both variables.
How closely can a link approach its signal-energy floor?
The ideal noisy-channel bound approaches a normalized energy floor as bandwidth grows, without a finite bandwidth optimum.
Next test or remaining limit
Measure total transmitter and receiver energy at fixed information rate as bandwidth changes, including circuit costs absent from the bound.
How does water recovery change ideal separation work?
Concentrating retained salt raises reversible work per product volume in the checked ideal-solution calculation.
Next test or remaining limit
Compare the ideal curve with measured osmotic pressures and activity-based calculations at controlled salinity and temperature.
Where does an ideal cell’s band-gap optimum lie?
Absorption, carrier energy and radiative losses produce an intermediate gap maximum in the supplied blackbody-cell grid.
Next test or remaining limit
Repeat the calculation under a measured illumination spectrum and compare with device data under matched optical and recombination assumptions.
When does a transport window throttle a network link?
A finite amount of data in flight limits ideal throughput through the window-to-round-trip-time relation.
Next test or remaining limit
Measure throughput while controlling window size and round-trip time, then quantify departures caused by loss and congestion control.
How does probe length change mismatch discrimination?
Consistent energy units change the supplied hybridization calculation, while melting shifts remain sequence- and condition-dependent.
Next test or remaining limit
Measure melting curves for matched and single-mismatch sequences at controlled salt concentration and strand concentration.
The linked reports retain their methods, code and limitations. Selection here does not establish novelty, universal validity or external peer review.
Methods, provenance & novelty reviewSource principles and review plans. The 81-candidate novelty register records pending assessments.
Novelty Investigation — 81 Research Candidates
81 focused dossiers across nine groups · 81 reviews pending · 0 completed assessments.
Open the review plan and candidate register
The supplied ChatGPT document proposes a novelty investigation. Its four named examples are suggested review targets; literature assessments have not yet been completed.
Search all 81 candidates, original dossiers and review history →
From question to evidence.
Research notes
Readable studies and literature summaries, with direct links to the underlying sources.
Data & methods
Tables, definitions and reproducible analysis, with provenance and units kept close to the numbers.
Visual explanations
Figures and diagrams that make the evidence easier to inspect, with captions, sources and limitations.
Methods beside the science
The reviewed science reports keep their assumptions, units, reproducible code and remaining tests beside the result.
Sources stay visible.
- Identify the study type, publication date and research cutoff. Label preprints, hypotheses and preliminary findings.
- Keep reported results distinct from our interpretation. Describe uncertainty and material limitations.
- Link to primary sources and explain data provenance, calculations and reuse terms.
- Disclose AI assistance and any review that actually occurred. Record corrections and updates.
Complete recorded release register62 recorded release routes including Set 8, Discovery07 and the I2 independent audit. Saved publication receipts; not a new scientific audit.
Separate URLs may describe revisions or overlapping collections. Do not add their counts as distinct discoveries.
- Set 8 · Practical Inventions400 proposals, 100 screened, twenty developments (six screened) and three supplied engineering studies; 22 original files retained. No prototype built.
- Discovery07 · The Black Swan Grid900 hypotheses, 300 screened, 30 editorial refinements (21 screened) and three supplied model reports; 26 original files retained. No new physical observations.
- I2 independent audit and open measurement testsIndependent mathematical checks, bounded primary-literature review, frozen observation protocol and proposed H0199/N0751 tests. No new physical observations.
- H0199 · saved-output correction48 saved rows reconciled; 12 errata and code limitations documented. Separate correction with originals preserved; no simulation rerun or physical test.
- Measurement Blind Spots300 hypotheses, 30 nested refinements and three supplied analytical/computational investigations. No physical experiments; corrections and all 27 original files retained.
- Earlier research libraryHistorical source archive/notes; publication and packaging verification does not establish scientific validity.
- Hypothesis notebookHistorical source archive/notes; publication and packaging verification does not establish scientific validity.
- Research data notesHistorical source archive/notes; publication and packaging verification does not establish scientific validity.
- ideas-20261030Historical source archive/notes; publication and packaging verification does not establish scientific validity.
- ideas-batch2Historical source archive/notes; publication and packaging verification does not establish scientific validity.
- data-labHistorical source archive/notes; publication and packaging verification does not establish scientific validity.
- methods-supplementHistorical source archive/notes; publication and packaging verification does not establish scientific validity.
- Original Theories Volume ISupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume IISupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume IIISupplied working hypotheses and proposed tests; not established scientific findings.
- Frontier IdeasSupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume IVSupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume VSupplied working hypotheses and proposed tests; not established scientific findings.
- Congestion hypothesis model testExact mathematics and batch-Bernoulli point-queue simulation; no empirical road claim
- Original Theories Volume VISupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume VIISupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume VIIISupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume IXSupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XSupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XISupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XIISupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XIIISupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XIVSupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XVSupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XVISupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XVIISupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XVIIISupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XIXSupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XXSupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XXISupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XXIISupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XXIIISupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XXIVSupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XXVSupplied working hypotheses and proposed tests; not established scientific findings.
- Storage chronology nominal-profile testNominal source-labelled profiles, not physically completeUTC chronology or actualFrench energy/storage sizing; source already includesduration, no whole-hypothesis refutation or noveltyclaim
- Original Theories Volume XXVISupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XXVIISupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XXVIIISupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XXIXSupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XXXSupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XXXISupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XXXIISupplied working hypotheses and proposed tests; not established scientific findings.
- Original Theories Volume XXXIIISupplied working hypotheses and proposed tests; not established scientific findings.
- Studies 1 and 2: compact three-level archiveFive conditional model analyses, with 4,200 supplied hypotheses and nested selections. Zero physical experiments reported; mathematical/model scope is not physical proof.
- Corrected pharmacogenomics coverage studyCorrected coverage analysis; see the published study for exact scope, sources and limitations.
- 900-series three-level source archiveSupplied prospective hypotheses, evidence-linked leads and focused dossiers. No physical measurements or completed proof of underlying hypotheses recorded.
- 900-series discovery reader, second editionAlternate reader of the same 900-series corpus; categories and associations are editorial browsing aids, not independent validation.
- Expeditions I and IISupplied mathematical/computational result reports; scientific programs not rerun during publication. Runtime environments and caches explicitly excluded.
- Expedition IIISupplied reports with correction, refutation and conditional-result labels preserved; no independent reproduction during publication.
- Expedition IVSupplied reports; inconclusive status preserved. No scientific programs executed or Blogger promotion during publication.
- 81-candidate novelty investigation registerReview brief and pending register, not completed novelty findings; same 81 source dossiers, not 81 additional hypotheses.
- Expedition VSupplied investigations and original files; no research programs executed during publication.
- Version 8 exploration workbookExploration items, not a 300-hypothesis result corpus. Original supplied categories retained.
- Version 9 partial framing snapshotPartial framing snapshot only; no completed study entries. Later content requires a separate revision.
- November 1 H034/H160 follow-up reportsH034 empirical question inconclusive: derivation/synthetic protocol only. H160 supplied conditional analysis with unresolved numerical check and unverified novelty at publication; separate corrections retained.
- H034/H160 independent bounded correctionsArithmetic/units correction and local conditional precision simulation. Not full FFT reproduction, physical proof, multitarget proof or novelty clearance.
- Further: eight developed directionsSupplied source-data reanalysis, mathematical derivations, simulations and proposed designs; negative findings retained. No new physical experiments; not independently reproduced during publication.
Earlier entry guide
Energy storage and chronological order
The current flagship engineering result, with data, code, limits and a defined validation path.
Read the flagship →Computations, datasets and code
Checked mathematics, conditional models and supplied results, each labeled with its evidence level.
Browse tested research →Questions awaiting investigation
Historical theory volumes and proposals remain available as preserved, searchable working archives.
Open the archive →New research · 3 November 2026
Three additive source-preserving bundles: Nov3 (600 hypotheses → 300 screened → 150 shortlisted → 15 near-proof dossiers), nov3pt2 (300 ideas → 30 proof attempts), and a separate public-record legal dossier.
No supplied program was executed for publication. The first two packages label their proof work as sketches, transplants or conjectures; novelty is unverified. The legal dossier is not scientific or medical evidence.
Read the collection note → · Download the preserved release ZIP →
Ten hypotheses, tested · 3 November 2026
I2 is the strongest metrology candidate (strict known-parameter test: forward model agrees to 6.4×10⁻⁵; the single-spot inversion is the bottleneck). P2 is the best low-cost test (illumination-side antisymmetry). H203 stays conditional: the CNEOS impact energy is a deterministic recalibration of the radiated energy (r=−0.995). P1 and P3 are corrected theory notes, not discovery claims. The round also covers H0199, H123, H153, H541 and H346, plus the four requested simulation suites (N0751, P2, H0199, I2). No new experiment.
Download the book · free · CC BY 4.0
Discovery with Machines: A Field Guide to Research with AI — twelve chapters on how research actually fails when a machine is doing the generating, and what to do about it. Illustrated, about nineteen pages, free to download.
Download
PDF edition — 60 pages, 24 chapters, 4 appendices, references and index.
EPUB edition — 32 sections with a full chapter contents list.
Text licensed CC BY 4.0. Corrections welcome and recorded.