June 26, 2026 ~1 minute minutes read Admin

Revisiting Choices: The Missing Science Skill

A blog post arguing that scientists should focus more on improving their decision-making processes than on acquiring more computational power.

Many scientists spend more time choosing which GPU to rent than which question to ask. That's a problem.

Science isn't a simulation. It's a sequence of decisions: what to pursue, who to collaborate with, which model to trust, who to hire. Yet we train students on methods, not on decision hygiene.

Every choice has an expiry date. The project that made sense six months ago might be dead on arrival now. The collaboration that felt promising? Check its half-life. Good teams don't just execute well, they revisit their assumptions before they expire.

"More compute" gets the headlines. But the real advantage is better judgment per CPU hour. That means building systems for reversible decisions, for killing projects early, for updating priors when data contradicts your hypothesis.

What if your lab had a decision audit? Not just what you chose, but why, and when you'll revisit it.