The real story in the google agents for science paper is that there is more ROI than ever for being a practicing scientist or engineer.

We need to figure out how to use AI without becoming dim: 
www.theintrinsicperspective.com/p/brain-drain

Over a decade ago I had the pleasure of coauthoring a perspective with the great Nick Jackson. Now we reprise the collaboration to discuss 10 problems in polymer chemistry that are plausibly solvable with emerging methods. For the framing device we are indebted to Mark Ratner: tinyurl.com/ytp5uzkm

Good information requires a lot of compute.

It seems intuitive that there is a Pareto front of model accuracy and interpretability. The trade-off is proportional to the irreducible complexity of the underlying mapping.