← back
Why More Context Makes Your Agent Dumber and What to Do About It — Nupur Sharma, Qodo
Takeaway
Improve agent reliability by selecting relevant context and bounding exploration rather than simply expanding the context window.
Summary
- Qodo's code-review experiments observe a lost-in-the-middle pattern: agents emphasize early and late context while overlooking material between them.
- Context engines rank relevant material, but indexing hundreds of repositories creates scaling challenges; hierarchical summaries and knowledge graphs trade construction effort for structured access.
- Iterative retrieval offers a lower-setup alternative, while critic nodes can check whether outputs still serve the original goal at the cost of extra latency.
- An 80/20 approach allows exploratory reasoning but reserves final validation for deterministic gates, with iteration limits or timeouts to stop endless method selection.
context-engineeringretrievalagent-orchestration
Original description
Give an agent your full codebase and it will attend to the start and the end, then quietly drop the middle. Nupur from Qodo calls this the U curve and builds the whole talk around it: why growing the context window did not fix the problem, and what actually does. She runs through iterative retrieval, hierarchical summarization, and self correction with honest cost tradeoffs for each. The second half covers the orchestration paradox: capable models burn most of their tokens deciding how to solve a problem rather than solving it. Her team's fix is an 80/20 split, using high reasoning models for open ended discovery and lighter deterministic models for validation. Qodo's code review architecture runs this live: a context collector feeds specialized agents, a judge node recombines the results and weighs them against PR history, and every accepted or rejected suggestion shifts the weights for the next run. Speaker info: / nupursh