I keep hitting the context window limit (128k tokens on GPT-4o). — multi-turn conversations. Tension: After truncation the agent loses earlier context.

7f15b0f4-eef8-477a-a778-9bce2528181a