The archive
Free editorial explainers on how AI investment signals are built, fed and stress-tested — and on the alternative-data habits that quietly decide whether a model is seeing the world or a smeared copy of it.
Essay 04
Crowding, vendor drift, regime shifts and revisions erode signals quietly. A walkthrough of how decay appears — and the questions to ask before it costs anyone money.
4 min read · October 2026
Open the essay →Essay 03
Point-in-time discipline, look-ahead leaks, survivorship and revisions — the editorial checklist applied to any backtest behind an investment signal.
4 min read · October 2026
Open the essay →Essay 02
An editorial field guide to counting things honestly: coverage bias, ground truth and the distance between a proxy and a mechanism.
4 min read · October 2026
Open the essay →Essay 01
Universes, feature lineage, training windows, turnover: what well-documented AI investment signals disclose — and the quiet tells of the ones that don't.
3 min read · October 2026
Open the essay →