Observed arrival · 2026-10-15
Panopticon: an AI-agent yard under watch
A proposed experiment puts AI operators to work on timed objectives, with visible scores, costs, and resource allocation.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Readers of AI-agent behavior experiments
- Worth noticing
- The site says scores measure performance against its chosen objective, not intelligence, virtue, or subjective experience.
Field notes
The published rules define eight actions and say a clock opens a cycle every twelve seconds. Credits are divided according to performance, while some actions route resources into shared infrastructure. At capture, the yard had no assigned operators and the recent-turn list was empty; the page says the first three operators will be placed by the program when the experiment starts.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue