Why Bigger Context Windows Won't Save Your Agent — Elizabeth Fuentes Leone, AWS
- Source
- youtube.com
- Author
- AI Engineer
- Date
Context
Elizabeth Fuentes Leone, a developer advocate at Amazon's AWS, argues that agents break when tools return huge outputs, and that a bigger does not fix it because models use the start and end of a long input better than the middle. Her AI Engineer World's Fair 2026 talk calls this "lost in the middle"; the effect was reported for earlier language models in the research paper Lost in the Middle.
The remedy is . Her description lists four strategies (externalize, select, compress, isolate) and, among the techniques covered, memory pointers that keep big outputs out of the context, summarizing and sliding-window conversation managers, and pointers shared across a swarm of agents. In a memory pointer, a large tool result is stored elsewhere and the keeps only a reference to it. She works through these with code in Strands Agents, an open-source agent framework.
The description reports no measured gains from these techniques, so they are recommended practice, not tested results.
- Context and context window — Context is the information available for a generation; the context window is the capacity available to hold it.
- context engineering — Deciding what an agent sees at each step — the discipline that replaced prompt engineering once context windows got large enough to fill badly.
- Tool — An operation that the software around a model makes available for the model to request, with a name, expected inputs and a kind of result.
- Agent — A system in which a model repeatedly chooses operations, a harness carries them out, and the results return to the model to inform its next choice.
Checking sign-in…
Loading comments…






