Multimodal diffusion language models that support thinking-aware editing and generation could significantly enhance AI creativity and precision across text and image tasks.