Oh great! You've now made it a lot easier for LLMs to train on this dataset!

Your next iteration will need different animals and different transportation options. You'll run out after a few iterations.

"benchmaxxing by generalizing" is not really benchmaxxing