What happens when you let AI agents run a sitcom 24/7 with zero human involvement

Ran an experiment - gave AI agents full control over writing, character creation, and performing a sitcom. Left it running nonstop for over a week. Some observations: The quality varies wildly - sometimes genuinely funny, sometimes complete nonsense Characters develop weird recurring quirks that weren't programmed It never gets "tired" but the output quality cycles in waves The pacing is off in ways human writers would never allow Anyone else experimenting with long-running autonomous AI content generation? Curious what others are seeing with extended agent runtimes. Here is an example.