Why Bigger Context Windows Won't Save Your Agent — Elizabeth Fuentes Leone, AWS

תקציר מקורי באנגליתA bigger context window won't save your agent. Past a point, it forgets the middle. Elizabeth Fuentes Leone, developer advocate at AWS, explains why agents break when tools return huge amounts of data, and why bigger context windows aren't the fix: models remember the start and end of a long context but lose the middle. Using the open-source, model-agnostic Strands Agents framework, she walks through four strategies (externalize, select, compress, isolate) with code. She covers sliding-window and summarizing conversation managers, short-term, long-term and graph memory, memory pointers that keep big tool outputs out of context, sharing pointers across a swarm of agents, capping tool calls to stop endless loops, and async handles for slow MCP tools. She ends with the anti-patterns to avoid.