Context Packing
How much of a codebase does an agent actually need?
Compare manual, semantic, structural and hybrid retrieval inside an 8,000-token context budget.
Work inside a limited context budget and compare manual selection, semantic retrieval and structural retrieval. The goal is not to fit the repository into the window. It is to find the information that controls the task.
Deterministic ranking over a fixed repository. No live model is called.
Related research
This archive contains independent research, experiments, architecture studies, and technical observations. Some findings are exploratory and should not be interpreted as peer-reviewed scientific conclusions unless explicitly stated. Architecture studies of third-party products are based on publicly observable behaviour, published technical information, and engineering analysis. They are not claims of access to private source code or infrastructure.