A tutorial on automating experiment design so scientists waste less time on guesswork

Someone has written a detailed guide to Bayesian optimization — a method that uses statistical models to predict which experiments are worth running and which ones to skip. In practice, this means researchers can run fewer experiments to find the same answers, especially in chemistry, materials science, and drug discovery where each experiment costs time and money.

This is a tutorial, not a new discovery, so it's primarily a signal about what knowledge is becoming portable enough to teach. The underlying method (Bayesian optimization) already exists and has been used in industry for years, but the fact that a detailed tutorial is now on arXiv suggests the technique is crossing from specialist tool to something practical researchers across chemistry and biology might actually use. The real structural change isn't the method itself — it's the accessibility. If this tutorial reaches lab groups that have been running experiments purely by intuition and trial-and-error, they might halve their experimental spend without losing quality. The threshold question: how many research labs currently use any form of principled experiment planning versus guesswork? That number is probably smaller than it should be.