Context_Headroom_Optimization

Compress long-running technical threads by pruning conversational noise, recursively summarizing large technical logs, and producing a load-bearing context pack (decisions/anchors/keys) that targets ~40–60% context-window reduction without losing critical signal.