I built a tree format for LLM code edits. Against a fair baseline, it only wins where structure matters

Automated excerpt

My first benchmark compared Nodast with the model rewriting the whole TypeScript file. On a 500-line file, changing one constant: Nodast with tools: 1,122 input tokens, 57 output tokens, 4.5 s About a hundred times less output. Against that fair text agent, on the same single-constant edit, Nodast tied.

Selected automatically from source text; not independently written or fact-checked. Read the original for full context.