HOW DASH USES CONTEXT ENGINEERING FOR SMARTER AI

Summary from the TLDR Dev newsletter

Dropbox evolved its Dash AI system from a traditional search tool into an "agentic AI" that can interpret, summarize, and act on information. Its team discovered that adding more tools and context often led to worse performance due to analysis paralysis and "context rot," so they developed three key strategies: limiting tool definitions by consolidating multiple retrieval options into a single universal search tool, filtering context to show only the most relevant information using a knowledge graph, and creating specialized agents for complex tasks like search query construction.