After looking at the PISA 2025 data, I noticed that the psychometric effects of the transition to a multistage adaptive design this cycle weren't accounted for at all. How strange, since ability-based item routing biases the naive Rasch linking they use.