It's not a very good cash hustle if people actually use it. From what I can tell, it's a one time $200 payment and you can use the LLM indefinitely. If the device costs $100 to produce and ship that's only $100 left to cover compute... forever.
It seems they want the compute to run on device, but that doesn't seem to be the case right now[0]. I don't know the specs of the device, but pretty sure it'd have to cost more than $200 to run an LLM locally.
edit: 128 gigs of storage and four gigs of RAM. For llama3-8b (the smallest llama3 model) you need at least 8GB of RAM.
0 - https://www.theverge.com/2024/1/18/24043490/rabbit-r1-ai-per...