 ISMB ECB Day 3 recap. Yesterday I spent the day listening to talks about synthetic data, used a large language model and visualization. The morning started with an excellent keynote by Ginny Jessi Kali, who received the overtone prize for her work on statistics and bioinformatics, especially related to synthetic null data as control. She talked about how problematic assumptions in RNA-seq analysis can lead to identifying wrong regulated genes and presented a method SCD to evaluate Disney and UMAP representations of SC RNA-seq data. Afterwards I went to the special session on large language models, where a good introduction led to a discussion about how these models are mainly reactionary, not doing deep thinking. We had a presentation about the acceptable use in publishing and reviewing, and then one about the impact on learning and evaluation. Next I went to BioVis. Here I learned about a method, FastNetVis, for kinase substrate network construction and visualization in 3D, and an invited talk on photorealistic molecular visualization using both VR and 3D printing.