Skip to content
How to ThinkIn the Age of AI
Principlev1PR-1068

Direct each optimization iteration at the single variable…

Direct each optimization iteration at the single variable identified as the current bottleneck, then measure results before changing additional variables, because simultaneous multi-variable changes prevent causal attribution.

Why This Is a Principle

Derives from Learning occurs when outcomes differ from predictions… (learning from prediction error) and Externalization Exposes Hidden Structure (externalization exposes gaps). This prescribes the scientific method applied to optimization: control variables, measure outcomes, learn from differences. It's the operational translation of PDSA.