Testigo

Testigo comes from a real payroll-system migration I did with an AI agent, where we lost weeks to errors that were there all along. It puts a migrated app next to its legacy system and leaves a record of what changed. For each record it captures both systems when the record opens and again after saving it without touching anything. Muse Spark reads every screenshot twice and proposes which old field matches which new one. Fixed rules then compare field by field: empty is never zero, amounts are compared by value, and when the two readings disagree the row stays pending for a person. The result is an acta, a short report where every value sits next to its screenshot and each row says what to compare. We tested it only on our own invented apps. It flagged a seeded defect (three empty fields saved as 0) with no false equalities, and found no changes in a migration made by Muse Code. A general AI agent given the same screenshots found the same defect, so Testigo's value is the fixed, repeatable question and the evidence trail, not better detection.