In MindFlow, I introduce a voice agent that can guide creators through the cloud-state by stimulating their divergent thinking. Throughout this machine-man brainstorming, creators are invited to perform craft-oriented actions involving repetitive hand movements, whose dynamics are recognized by means of computer vision. Then a machine-learning algorithm analyzes the hand movements and explores the thought-map of the creator. It does this by identifying focal connections between mind-states and making, such as moments of high-focus, typically accompanied by stopping points in the gestures, and key points, where creators tend to explain themselves both verbally and by hand gestures.