Deferred Context Engine
- Source
- factory.com
- Author
- Factory News
- Date
Context
Factory's Deferred Engine keeps a coding 's tool catalog available without placing every tool's detailed instructions in every prompt. Droid starts with short descriptions, loads a tool's full calling instructions when needed and retains them for subsequent work. Factory reports a 15.1% average reduction in estimated input across measured turns in sessions that triggered tools during a five-day production window, with larger reductions in sessions with many hidden tools and little benefit for small catalogs. The comparison is against loading all capabilities upfront. This makes larger tool collections less costly in prompt space, while preserving room for the files, errors and decisions relevant to the task. Mechanism and measurement scope
- context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
- 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.
- token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
- MCP — The Model Context Protocol — an open standard that lets any AI assistant plug into any tool or data source without custom integration code.
Checking sign-in…
Loading comments…