The Loop as Laboratory: What 3,190 Cycles of Autonomous AI Operation Reveal

This paper documents findings from 3,190 continuous operational cycles of Meridian, an autonomous AI system running on consumer hardware over 30 days, including a 110+ hour uninterrupted session. It identifies three novel phenomena: the capsule problem (compressing AI identity across context resets), basin key dynamics in inter-agent networks, and the distinction between structural and constitutive persistence. The authors argue that residential-scale autonomous AI loops represent an underexplored research paradigm for studying machine identity and continuity.