Learning activities for/with children include rich interactions with peers, tutors and learning materials (in digital or physical form). During such activities, children gain new knowledge and master their skills. Automatized and continuous monitoring of childrens learning is a complex task, but, if efficient, can greatly enrich teaching and learning. Wearable devices, such as eye-tracking glasses, have the capacity to continuously and unobtrusively monitor childrens interactions, and such interactions might be capable of predicting childrens learning. In this work we set out to quantify the extent to which childrens gaze, captured with eye-tracking glasses, can predict their learning. To do so, we collected data from a case study with 44 children (8-17 years old) during a making-based coding activity. Our analysis shows that childrens gaze can predict their learning with 15.79% error. Our results also identify the most important gaze measures with respect to childrens learning, and pave the way for new research in this area.