📰 Dev.to · John

A pre-registered, randomized experiment did everything right — frozen metrics, appended amendments, a falsifier that blocks forcing a verdict on too little data

A submission gate required beating a proxy 24x noisier than its real target, and a near-zero pass rate was a symptom, not proof the gate worked.

Building a travel app on Korea's government open-data portal, every researched guess about operation names, fields, and approval scope was wrong until a live ca

Our A/B assignment used a cryptographically fair coin and was still unanalyzable — nothing joined the random draw to the task it dispatched, so no code could ev

Other project idea I recently decided to build in public was inspired by the NASA Exoplanet Query...