Every single day, people around the world take numerous photos with their smartphone cameras and receive lots of photos through applications such as WhatsApp. However, only few of them are actually meaningful and worth reviewing. Kioku aims to filter out these photos, the ones which carry value, from your film roll. The photos that are recommended by the system are selected through the concept of machine learning, where “the seasons”, and “your personal calendar” are used as input factors. The concept allows you to rate the presented photos by either smashing the interface (bad), or stroking it (good). The system responds to this by changing its shape accordingly. Over time, Kioku will, based on your ratings, learn which photos you like to be reminded of. As Kioku gets to know you better, it will understand which memories match both your agenda and your preferences. The interaction with Kioku was inspired by the idea of communication between the user and the artefact. In order to apply this idea, Kioku starts to breathe in order to indicate that it has found a memory. The user can then gently stroke Kioku to indicate that they want to review the memory. If the user likes or dislikes the memory they can indicate that. In case of a dislike, Kioku shrinks itself, almost indicating that it is sorry. In case of a like, Kioku looks proud and becomes a bit more inflated. In order to be able to create this shape changing interface, a gear mechanism was developed. For the machine learning aspect, k nearest neighbour (kNN) and content based filtering were used.