This paper studies how short-term variations in fine particulate matter (PM2.5) pollution affects cognitive performance across tasks of varying complexity. While prior work shows that pollution impairs performance in highly demanding cognitive settings, it remains unclear whether these effects extend to simpler tasks. We examine this question using data from official Rubik’s Cube tournaments in the United States and India. Solving different cube sizes provides a natural proxy for task complexity, while solving time measures cognitive performance. To identify causal effects, we exploit exogenous variation in local PM2.5 generated by wind direction. We find that PM2.5 pollution has negligible effects on simple tasks but significantly slows performance on complex ones for tournaments in the United States. In India, where baseline PM2.5 levels are substantially higher, we find similar effect patterns but none of the effects are statistically significant. We show that this pattern is explained by diminishing marginal sensitivity to short-term PM2.5 shocks as baseline PM2.5 pollution levels increase. Our findings provide causal evidence that the cognitive costs of PM2.5 pollution depend critically on task complexity.