Writing
The Science Behind Causal AI
White Papers & Technical Analysis
All posts
Everything we have published.
Sep 28, 2026
How we got to SPARC-fast
Three versions of our causal discovery engine: the score we had to lose, the model we chose not to train, and how SPARC-fast won the speed back.
Jan 12, 2026
Why Most AI Explanations Are Fake
SHAP and LIME tell you what a model learned from historical data - not how the world actually works. Here's why feature importance scores can't answer causal questions, and what enterprise decision-makers actually need from AI.