Show HN: Parsimony – an open-source benchmark for how much code AI models add

Parsimony is an open-source benchmark that measures how much code AI models add when solving programming tasks, addressing a gap in existing benchmarks that only track solve rate and cost. It counts coding units (syntax elements like statements and calls) rather than lines, penalizing excess code while crediting deletions, and ranks models by the lowest mean net units added on solved tasks. The tool includes a leaderboard and allows users to compare AI models across different metrics while acknowledging limitations like the correlation between code volume and task completion.