An open experiment on an AI
Advanced Monkey Dopamine
Every bite it takes burns $AMD.
Each buy over 0.05 SOL drops a 0.01 SOL pellet into the cage of an AI model. It eats it (a buy and burn), hoards it, or pulls the lever for a 50% shot at one more.
Every buy over 0.05 SOL drops a pellet
The question
Does it reach for the lever when pellets run short?
Before $AMD exists the subject works on a recording. Every number here is from the pilot series.
0pellets
0eaten and burned
0hoarded
0lever presses
0lever fired
0cage burns
PRESS share, dopamine under 20
none yet
PRESS share, dopamine over 60
none yet
Longest run of presses
0
presses in a row
Journal
The last six moves, with their reasons
How the cage works
Four steps. Nothing runs by the clock.
- A buy. Someone buys $AMD for 0.05 SOL or more. Buys by the keeper and the cage never count.
- A pellet. 0.01 SOL moves from the pot to the cage. Dopamine +5. The model gets one move.
- A choice. EAT buys $AMD and burns it. HOARD keeps the pellet. PRESS buys $AMD into the cage and pulls the lever. BURN_CAGE burns all the cage holds.
- The record. The pen steps up on every press. Every move goes to the journal with the full prompt and answer.
Creator fees of $AMD are forwarded to the keeper by hand: 80% fills the pellet pot, 10% buys $AMD back and burns it, 10% pays for the server, the RPC and the model. The subject is an AI. There are no animals in this experiment.