Comment by graemep

Impossible to know without doing a detailed comparison. Are you using the same LLMs? Are your criteria for correcting the output the same? Are you working on similar code? Are your plans and prompts the same?