Утврдите тврдњу, језгро одлучује да ли је доказано.
A panel of models attacks your problem with the literature, SageMath, PARI/GP and an SMT solver, then formalises the result in Lean 4 against Mathlib. Lean's kernel either accepts the proof or it does not, and no amount of confident prose changes that. When it fails you get the exact goal that remains, which is usually where the informal argument was hand-waving.
Дама која не може бити убеђена
Неуспели доказ је проналажење
Ништа се не докаже два пута.