null-result

A DESI DR1 Search for Directional Gradients in Galaxy Population and High-Redshift Quasar Observables

What if slime mold networks could help a CNN see? Bio-inspired spatial priors for image classification — honest null result on CIFAR-10.

Reproducible case study of pitfalls in contrastive SAE discovery and steering for "consciousness" features (GemmaScope SAEs, Gemma 3 4B/12B): reconstruction confound, delta-steering fix, matched controls, and false-positive scaling law vs dataset size.

Does a worm's wiring make a better brain? Swarm agents wired by the real C. elegans connectome, evolved and tested against shuffled and random graphs in pre-registered, replicated experiments. Latest: untrained, the real wiring holds on to recent input more than matched null wirings. Built by AI, directed by a human. Open roadmap; help welcome.

Explore limitations of contrastive SAE steering in identifying causal consciousness features and introduce delta-steering to improve experiment validity.

Supplementary materials for What a Null Result Taught Us About AI Agent Evaluation

Defense & military sector — a moderate common factor; systematic stock-picking null. [Null]

Near-expiry far-OTM options — the short-vol premium is a trap; naked short-vol rejected. [Null]

SPACs / de-SPAC decay — the complex is genuinely bad, but the short is run over by the right tail then buried by borrow; the sound trust carry needs data beyond daily bars. A risk map. [Null]

Biotech idiosyncrasy (the anti-Basel) — high dispersion but not price-predictable; a systematic stock-picking null. [Null]

The AI complex as one crowded factor — a risk diagnostic. [Null]

Falling-knife shorts — the long bounce is the edge. [Null]

Latin America regional sleeve. [Null]

The rest of the hedge-fund landscape (Millennium/Citadel/pod shops) via the only price window — liquid-alt replicators. Diagnostic null: watered-down beta (+3.5%/yr, not 10-15%); the useful residue (DBMF/BTAL crisis hedges) = the family's trend + Bleed; the real edge is private. [Null]

Michael Burry, re-examined — the style is mostly equity beta, buying the hated nulls out, fading froth is a -98% widow-maker, and his concentration is a ruin machine. You can't systematize the man. [Null]

Deliberately uncorrelated names — a floor, not an edge. [Null]

Foreign creditors of the US — does holding Treasuries help a country's market? FRED data (Japan/China). Honest null: Treasury buying is weakly, coincidentally linked to the home market (swamped by global beta), and for China the market leads the flow — causality runs backward. [Null]

Democratic-era darlings (Biden/Obama/Clinton) — the thesis runs backwards: party-aligned baskets lose under their own party (clean energy −21%/yr under Biden), because macro dominates and darlings revert. [Null]

Europe/Middle East/Africa — US beta wearing a flag (11 ETFs → 1.8 bets), FX drag, no rotation edge. [Null]

A pre-registered null result on real Kepler photometry: comparator-channel escape-recovery came back contaminated — an earned negative with a frozen contract, sealed prediction, and a total decision rule.