Humane Ai Pin collapsed, but the rabbit is still kicking. The company has published Post and video on the blog Today, showing off the “generalist agent on Android”, slowly controlling applications on the tablet in the same way as Rabbit claimed that his R1 device was over a year ago. (It could not and could not.) The work is based on Lam Playground, the rabbit “generalist web agent” launched last year.
Engineers do not exploit the R1 rabbit to demonstrate at all. Instead, they enter their demands in the line on the laptop, which translates them into actions on the Android tablet. These are things such as finding a movie on YouTube or locating the recipe for whiskey cocktail in the cocktail application, collecting ingredients, and then adding them to the Google Keep food list. At some point they ask for a logical game 2048 And find out how to play, what he does, though slowly.
The model usually does things they ask for, sometimes good, and sometimes with weirdos, such as sending a poem by WhatsApp, one message at once instead of in one block. One of the engineers wonders if they should ask him to exploit a break on the hint, but they don’t come back to try again.
AI Rabbit agent is clearly working, just like since R1 launched almost any of the possibilities that the founder and general director Jesse Lyu presented in January 2024. Rabbit constantly introduced updates, such as the possibility of training his AI agents to perform specific tasks or inclinations to modify your own interface. The examples he presented today are “only the basic loop of the action, which is completed by the Android agent”, according to the post on the Rabbit blog. The company promises to share more about the “upcoming inter -system system of many agents” in the coming weeks.