Factory describes a staged approach to giving coding agents relevant information. An initial repository overview supplies structure and build instructions. Search identifies candidate files, after which the AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → retrieves particular files, lines or changes as needed. Its architecture account also incorporates documentation and operational information outside the repository.
This addresses a specific weakness of retrieving similar text alone: a coding task may require following relationships between several parts of a system. Search provides starting points, while targeted inspection supplies the connections. The design treats the model's context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition → as working space to fill selectively as the investigation develops.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
