The Null Institute / Publications
Research Notes
Short methodological observations, exploratory findings and open questions. Verification pilots are listed separately to prevent duplicate archives.
12 records
VA Psychedelic-Therapy Readiness: The Clinician Hours Behind 1,000 Courses
A reproducible contact-hour model: the staffing behind 1,000 annual MDMA-assisted therapy courses, and what VA still needs to disclose.
US First Births Rose From 2019 to 2025. Total Births Fell.
In the 2025 CDC birth records, first births were up from 2019 even as total births fell. Our independent count shows where the decline occurred, and why a Census denominator change matters.
The 1960 Doomsday Equation: Reproducing the 2025 Forecast and Its Sensitivity
Independent reproduction of the 1960 von Foerster population equation against the UN 2025 estimate, including a time-coordinate and singularity-date sensitivity check.
How Wrong Is the Inverse-Square Shortcut for LED Panels?
A reproducible reanalysis of four LightLab irradiance reports: the 24-inch laboratory measurements exceed a simplified point-source extrapolation by 7.9–10.1 times; four reported means independently reconstructed from 196 readings.
When Better IT Agents Cost More Per Run: An ITBench Cost Recalculation
An independent recalculation of IBM's 35-incident ITBench comparison finds EoG costs more per attempted diagnosis on recorded tokens for three of four models at an illustrative 4:1 token price ratio, but less per point of majority@3 entity F1 for all four. Break-even price thresholds, source data, code and limits.
Norway’s Wealth Tax: What Six Years of Household Data Reveal
In Norway, assessed wealth tax rose 72.8% between 2021 and 2024 while measured household net wealth rose 11.5%. The richest quarter consistently accounted for 98–99% of assessed wealth tax, a six-year recalculation from official SSB data.
When Two Biological Age Clocks Disagree: What Happened to 686 People?
Independent NHANES reanalysis: two biological-age clocks flagged 686 different people; ten-year mortality differed by 10.2 points, or 6.3 after weighting.
A Reversed Trial Entry Weakens a Heart-Failure Fluid-Restriction Meta-Analysis
Reconstructing seven estimable mortality trials, we find that reversing six Holst deaths changes pooled RR from 0.544 (p=0.0345) to 0.681 (p=0.2408). Holst exclusion, four leave-one-out tests, and modified Hartung–Knapp inference also cross the conventional significance threshold. Includes data, code, and limitations.
Four of Five Data Center Ban Votes Followed Earlier Local Restrictions
Independent cross-check of five November 2026 data-center ban votes: at least four municipalities had already enacted restrictions before Election Day (4/5 = 80%). Sources, classification rules, and limitations.
In Female Mice, a Longevity Model’s Strongest Signals Did Not Track Lifespan Gains
A five-intervention recalculation found a negative correlation (Spearman ρ = −0.789; exact p = 0.15) between reported female lifespan gains and the number of tissues distinguished by a longevity model. The observed pattern is descriptive, not proof of predictive failure.