video1.0<iframe src="https://www.loom.com/embed/4a819797f1c3429f90445d16f8bf7d00" frameborder="0" width="1658" height="1243" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>12431658Loomhttps://www.loom.com12431658https://cdn.loom.com/sessions/thumbnails/4a819797f1c3429f90445d16f8bf7d00-c3c1a930dd1e9c3f.gif99.029333This Loom shares a structured analysis of how different memory approaches affect computer use agents interacting with web pages. Ricardo compares flat memory versus graph memory and a structured graph memory setup, focusing on success rate and the average number of steps needed to create actions on a site. He reports that flat memory improved performance versus the others, while graph memory showed little difference, with room to test better graph tool usage. He concludes that the work is constructive and looks forward to improving scoring, exploring different graph traversal methods, using stronger benchmarks, and testing in real-world experiences.