What should they emphasize instead?

Because the blog post will be weeks or months away given my current workload, here's a thread.

First, your whole analysis from raw data to final figures and tables should be software generated in a fully reproducible way. Great way to avoid errors and find them if they crop up.

I’m just brainstorming, but in the AI to Neuro direction, you could apply constraints to obtain biologically plausible normative models during tasks like motor control, then use these models to develop tools to analyze both ANNs and BNNs in terms of latent spaces/representation spaces