Drift and selection in LLM text ecosystems

Create a simulator that models text ecosystem drift using n-gram agents. This tool would help researchers quantify the long-term impact of AI-generated content on datasets.

Suggested repo: drift-sim

"Visualize the collapse of your training data pipeline before the next LLM generation cycle."

Estimated effort: 50h