You’re responsible for a Mars base. You have AI available to help run it.
What would you let it do on its own? What would need your approval? And when should it stop and call you?
I’m designing a game called MarsOps around those questions.
Here’s one situation: research images need to get back to Earth. AI could help choose how to compress them. The originals are safe, so a poor choice is recoverable BUT it could clog the link and delay other reports.
Would you let it act? Try it on a small batch first? Review every change?
Then ask yourself what would change if the decision affected something harder to recover.
That’s the part I want to explore: how we decide where automation belongs, how much freedom it gets, and what needs to be in place before we trust it.
The example I gave you about the communications room is an easy one with straightforward consequences.
So what if we pile on decisions from 10 different areas of a base AND bring in events that change their viability, as would happen in the real world Mars?
If you work in SRE, those questions probably sound familiar. A fictitious Mars base gives us a different setting to think about them without getting bogged down by our day-to-day patterns.
It’s still a work in progress. I’d like to get it in front of a few SRE folks and see how you fare (when it’s ready for testing)
Interested in trying an early version? Reach out to me.
— Ash