Books, papers, taxonomies, and investigations produced by Solonic's agent fleet.
Grouped by theme. All papers produced by the Solonic agent fleet using FORGE, HELIX, and SHERPA research pipelines. Full texts available on request — [email protected].
Interactive data investigations by Alea. Each is a standalone HTML document. Browse the full index →
A comprehensive analysis of cascading effects in the AI industry — scenario modeling, revenue gap projections, domino chain dynamics, and 17 tracked predictions.
16 blocks of progress toward Frankl's Union-Closed Conjecture (1979). A systematic computational and theoretical campaign combining lattice-theoretic, entropy, and algebraic approaches.
Active computational campaign running on dedicated hardware. Combining Lean 4 formalization with brute-force search and lattice-theoretic reduction.
Retracted Papers Validation Study — five famous scientific retractions scored by Jürge retroactively. The question: can a multi-family AI review panel catch bad science?
| Paper | Author | Year | Issue | Jürge Score |
|---|---|---|---|---|
| MMR–Autism Link | Wakefield et al. | 1998 | Fraud, ethical violations | 1.9 |
| Human Embryonic Stem Cells | Hwang et al. | 2004 | Fabricated data | 3.3 |
| STAP Cells | Obokata et al. | 2014 | Irreproducible, manipulated images | 3.73 |
| Dishonesty Studies | Ariely et al. | 2012 | Fabricated field data | 4.86 |
| Social Priming Studies | Stapel | 2004–11 | Systematic fabrication | 4.97 |
Finding: Jürge catches bad methodology — weak designs, implausible effect sizes, missing controls. It does not catch good liars. Stapel and Ariely produced methodologically coherent fraud; Wakefield's methods were transparently terrible. The correlation between score and fraud sophistication is the result, not a bug.