A week with Cue, Bot and Muse as personal assistants
After a week of testing Cue, Bot and Muse, the author finds them decent at task execution but far from managing errands and long-term plans as personal life assistants, largely because they lack a world model.
After using @CueAgents, @bot, and @Muse for a week, I’d say they’re reasonably capable if all you need is task execution. But as personal life assistants that manage everything from small errands to major plans, they still fall far short. I believe the biggest gap is that current AI agents lack their own World Models, which makes long-term planning and reasoning difficult. This doesn’t require any changes at the model level; it only requires introducing a world model at the architectural level.