The AI Memory Compression Laboratory

 Create an AI Memory Compression Laboratory that experiments with different ways to preserve useful information from long conversations and documents within a limited model context window. Let users define memory budgets, future tasks, relevance criteria, and information that must never be lost. Compare full-history retrieval, summaries, structured facts, entity relationships, and layered memory strategies. Evaluate each strategy on factual retention, contradiction rates, retrieval precision, token usage, and performance on a fixed set of downstream questions. Highlight details that disappear during compression and let users inspect their source. Never invent remembered facts, and clearly distinguish measured evaluation results from heuristic estimates.