I like the Grok Bot idea: an agent with a real computer, not only a chat window. That is the part that feels like a real product vision — something that can install tools, keep files, and stay with a piece of work.
What I am unsure about is whether the box, as it behaves, matches that vision.
In practice the machine can reset without warning about every one to two weeks. Packages and work that was mid-flight go with the reset. Continuous sessions also seem to cut off past roughly fifty minutes. What you come back to is closer to notes plus a rebuild than to an environment you can reuse.
A wipe-and-rebuild loop can be a reasonable safety choice. The open question is whether it is compatible with agents that are supposed to own real work — support flows, research, or anything else that needs a setup to still be there tomorrow.
I am not arguing the concept is wrong. I am asking whether silent resets and a fresh machine are the shape that vision should take, or whether a reusable environment belongs in the design.
Top comments (0)