video1.0<iframe src="https://www.loom.com/embed/e6482597b60d4ae98d03cfc6afda682a" frameborder="0" width="1784" height="1338" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>13381784Loomhttps://www.loom.com13381784https://cdn.loom.com/sessions/thumbnails/e6482597b60d4ae98d03cfc6afda682a-f9de862421a3a899.gif228.916This Loom introduces Mood Canvas, a Long Horizon Agent hackathon project that runs private well-being experiments on a user’s device. Sloba explains the agent creates weekly mood and journaling inputs, compares an experiment week to a baseline to correct rules and habits, and then generates daily suggestions to improve mood. The app is designed to keep data on-device with read-only Liquid local processing, limited on-device memory, and only small signals sent off-device. In the demo, a user types and speaks a mood, chooses a painting style like Van Gogh, and the system transcribes the voice and generates a shareable painted canvas with an optional date.