but even then it is not so trivial. Yesterday I gave DeepSeek a simple diophantine equation and it got it wrong 3 times, tried to correct itself and didn't end on a correct solution, but rather lied that the final solution is correct.