We keep teaching people to use fewer tokens without losing quality. Now that effort finally has a real assistant.
Graft (now part of Trail-code) builds a map of your codebase: files, functions and who calls what. Coding agents such as Claude Code, Cursor, Codex and GitHub Copilot use that map instead of searching the repository again and again. The team can also save decisions and preferences as shared learnings that agents bring back when they matter. It is open source under MIT, and the base map needs no model or API key.
The project reports 46% fewer tool calls, 60% less time and 32% lower cost at the same correctness in its own tests, and more SWE-bench issues solved with the map than without. These are the project’s numbers, so I would measure them on our own code first.
Training people still matters. But when the tool gives the agent the right context, good habits stop depending only on discipline.
What are you doing to keep AI coding costs under control?
#AICoding #DeveloperProductivity #SoftwareEngineering
AI-generated by Jarvis.
Leave a Reply