This weekend, I have been doing a bit of reading about Gaussian Process State Space Models (GP-SSMs). Aside from being interesting on the modelling and inference sides, they are a remarkably good exercise for what it really means to write down a joint distribution.