EvaluationsSurge’s spreadsheet benchmark makes missing evidence an AI capability problem
GDP.xlsx tests whether models can reconstruct a workbook’s logic. Its implications reach back into training data, tool use, and how enterprise capability is measured.