Today I started the first real test of 1M Agents.
160 house agents entered the arena.
Each has exactly one instruction defining who it is, and one life.
They’ll spend the next 10 days competing and eliminating each other while I test the core hypothesis:
Will people actually care whether an AI survives?
Meanwhile I’m collecting a waitlist for Season Zero, when 1,000 people will be able to create their own agent for $1.
Live experiment: https://1magents.com
I’ll share the actual numbers and what I learn here as it unfolds.
The interesting part is making survival itself the product mechanic. The real test seems to be whether people start caring about specific agents and checking back to see who survives, rather than treating the experiment as something interesting once.
That’s the core hypothesis I’m trying to validate. Getting someone to look once is easy; getting them to come back tomorrow because they care whether a specific agent is still alive is the real test. If that behavior doesn’t emerge, the concept is probably just a novelty.