Comment by vidarh

The comparison would be apt if the LLM was trained on your codebase.