This week our software side reached a milestone we had been building toward: we started training the machine learning model that turns biometric signals into a read on stress, using the datasets we had carefully selected in earlier weeks. It was the first time our core idea stopped being a diagram and started becoming a working classifier. Alongside the model, we finalised the first complete version of the prototype, closing the gap between a breadboard and a device that could actually be worn. We also presented our progress to colleagues in the course, walking them through the technical details and the advancements of the past weeks.