A STAT+ write-up highlighted how AI progress in drug development may be running ahead of practical expectations in drug metabolism. The piece pointed to competition results suggesting that bigger models do not automatically outperform smaller systems when the task is predicting whether candidates activate PXR, a key nuclear receptor controlling metabolism of a large share of marketed drugs. The central operational challenge: many development programs identify PXR activation late, forcing late-stage reformulation. The article argues that reliable early prediction of PXR activation could reduce rework, mitigate drug–drug interaction risk, and improve candidate selection earlier in discovery. The update situates metabolism-focused AI as a high-impact use case for model development, but emphasizes that performance gains are likely task-specific rather than simply scaling-driven.