Sol Innovations — UET-NET Research
Founder and Principal Investigator
Anthropologist and U.S. Air Force veteran. Directs an independent research program that tests cross-domain correlations — economic, geophysical and solar — against public data, and keeps a public, cryptographically anchored record of every forecast the program makes.
Human societies respond to environmental change. The rains drive the wildebeest migration across the Serengeti on a regular schedule; human responses to the same class of forcing are real and far less direct, because human behaviour is more complex. They run through harvests, prices, migration, public mood and conflict, and they unfold over years to centuries.
This is an anthropological question asked with quantitative tools, and it has an established empirical literature: climate variability has been tied to war and population decline (Zhang et al., PNAS 2007), to the rise and fall of European societies over 2,500 years (Büntgen et al., Science 2011), and to conflict across 60 studies (Hsiang, Burke & Miguel, Science 2013). This program asks the same question of solar and geophysical forcing, arrived at independently through cross-cultural study of calendrical and sky-observation systems, and tests it at civilizational scale.
Scale determines the instrument. A month-to-month correlation against a market index can rule out a fast, direct link, and several of ours have been ruled out that way and published as refuted. It cannot test a slow, mediated one. Four preregistered tests, at the eleven-year scale and at the Gleissberg scale, have now run; none supports the hypothesis as stated, and each result is published with its script. The program keeps the question and reports the answers.
Hypothesised relationships between public economic series are tested and reported whether they hold or fail. Since September 2026 every pair is judged on period-over-period changes; the level coefficient is reported beside it for comparison.
Data: FRED (Federal Reserve Bank of St. Louis), monthly averages
The central test, at the scale of the thesis, run exactly as registered. Hypothesis: crisis onsets cluster in the declining phase of the solar cycle at a rate above chance. Result: banking crises (Reinhart–Rogoff, 70 countries, 1800–2016) show no clustering; sovereign defaults show a 12–16% excess in the declining phase that is significant only when every country in a default wave counts as a separate event, and falls to p ≈ 0.05–0.07 when each onset-year counts once. The two sources disagree, which the registration defines as list-dependence, so the hypothesis is not supported. A follow-up registered the same day, with the onset-year as the unit and a permutation test for wave clustering, found the same tilt (ratio 1.14, p = 0.053) and is not supported either. A third registration, run on two independently compiled chronologies (Asonuma–Trebesch monthly defaults, 1970–2019; Bank of Canada–Bank of England, 1960–2019), returned a ratio of 0.99 (p = 0.59): the tilt does not replicate and is recorded as refuted at the eleven-year scale. The thesis at its own scale was then registered and run the same day: Gleissberg-envelope phase (Usoskin 2021 radiocarbon reconstruction) against 2,433 conflict onsets in the Brecke catalog, 1400–1849. Descending limbs carry fewer onsets per year, not more (ratio 0.91, permutation p = 0.75); the highest solar peak was followed by the lowest onset rate; descending limbs are not colder in PAGES 2k. Refuted in its tested forms.
Data: SILSO cycle minima and maxima (Royal Observatory of Belgium) · Reinhart & Rogoff crisis chronology (HBS Global Crises panel) · NBER business-cycle dates
Continuous ingestion of planetary Kp, F10.7 flux, solar wind and magnetometer band power. Routes return an explicit failure state when a source is unreachable; no substitute value is displayed as a measurement.
Data: NOAA SWPC (3-hour Kp, daily F10.7) · HeartMath GCMS magnetometer network (hourly band power, ~2-day lag) · USGS seismic feed
Every forecast is logged as a structured call with its resolution criteria, then graded. Each call is hashed (SHA-256 over canonical JSON), the hashes are combined into a Merkle root, and the root is written to the Base blockchain, so no call can be edited or removed after the fact without detection.
Data: Call ledger and anchor manifest published at /record
Cross-cultural study of calendrical and sky-observation systems as records of environmental cycles. Each published claim carries an evidence tier (established, contested or speculative) and the principal objection is printed beside any contested claim.
Data: Published archaeological and historical literature; astronomical positions computed with astronomy-engine (VSOP87)
The genesis anchor covers 66 forecast calls. Their Merkle root was written to Base (chain id 8453) in block 50,553,167 on August 28, 2026, signed from a hardware wallet. Any call can be verified independently: canonicalize it, hash it, confirm the hash in the manifest, then confirm the root in the transaction calldata.
A September 2026 internal audit found that correlations published as validated had been measured on raw levels. Re-measured on changes, four of seven did not hold, including the result previously presented as the program’s novel finding (−0.89 on levels, −0.04 on changes). The figures were corrected at the source, in the public API and on the public pages, and each withdrawn result remains listed with its measured value.
Tests whose parameters could be tuned after the fact are registered in a dated document before data is touched. Revisions require a new dated document; both remain on file.
Validated claims use publicly available series only. Replication scripts are kept with each audit, and independent replication is invited at research@uet-net.com.
Claims are classified as empirical, in research, not validated or framework. Hypotheses that fail their thresholds are published in that state alongside the ones that pass.
The program is computational throughout: ingestion and normalization of public time series, correlation and lag analysis on differenced data, replication scripting, Merkle-tree construction for the record, and literature and citation verification for the comparative work. AI assistance is used for code, analysis scaffolding and source checking; results are judged on the statistical tests, and every validated claim replicates from public data without it.
Research dispatches are published at civscience.substack.com. Platform overview: About UET-NET.