Teun van der Weij comments on [Paper] AI Sandbagging: Language Models can Strategically Underperform on Evaluations