The Rabbit R1 companion device uses a "Large Action Model" to understand software interfaces, to help you actually accomplish tasks without launching apps
This is a really interesting new product, launching at $199 and available by March/April this year.
Featuring sleek retro-futuristic industrial design by Teenage Engineering, the Rabbit R1 'companion' has a rotating camera, a scroll wheel and clicky side button, far-field mics, a touchscreen, a rear speaker, a USB-C port and a SIM card tray (though it's not a phone, per se).
The Rabbit OS is really kind of a "super app" that can recognize UI elements in interfaces and thus learn to navigate them on it own to accomplish tasks.
You give it logins to all of your other apps and services, and thus can accomplish actual tasks from commands, like "book me a flight to San Francisco, plus a hotel and Uber when I get there" - whereas current iterations of large language model (LLM) AIs can really only just search the web and offer you links (when they're not hallucinating information that doesn't exist...)
To me, this feels like what API services like Zapier and IFTTT (If This, Then That) were supposed to do, without the need to manually wire up actions yourself.