Type: url
Original: https://arxiv.org/pdf/2512.07921
Status: generated
Summary
A paper presenting DeepCode, an open-source agentic coding framework (blueprint distillation, stateful code memory, retrieval-augmented knowledge injection, closed-loop sandbox verification) for synthesizing a complete code repository from a document such as a scientific paper, reporting it surpasses commercial coding agents and PhD-level human experts on the PaperBench benchmark.