ResearchResearch
FastContext: Training Efficient Repository Explorer for Coding Agents
The research paper behind the widely shared 56.2% tool-turn and 46.5% token figures. This is not a Jev evaluation.
Microsoft / Shanghai Jiao Tong University
Find the original file, then read the claim in context. This library keeps official documentation, independent analysis, research papers, and working notes clearly labeled.
The 12-page Jev engineering note has circulated as a founder release, but its cover says it was independently compiled and is not endorsed by TypeSafe. Our Jev Engineering for Coding Agents PDF guide includes the download, ten harness ideas, a practical walkthrough, and a check of the claims most often repeated with it.
Resources are grouped by format. Labels describe provenance, not an endorsement or quality score.
The research paper behind the widely shared 56.2% tool-turn and 46.5% token figures. This is not a Jev evaluation.