Check that the sum of two odd numbers is even: 3+5=8, 7+11=18, 101+999=1100. Every example works, but there are infinitely many odd numbers, so examples can never finish the job. A proof replaces "these cases" with "any case". Write any odd number as 2a+1, where a is an integer, and another as 2b+1. Their sum is 2a+2b+2=2(a+b+1), which is 2 times an integer, so it is even. One calculation covers every pair at once.
Each step must follow from earlier ones by a rule you could name: a definition (odd means 2a+1), algebra (collecting terms), or a result already proved. A step that sounds right but has no such reason is a gap, and a proof is only as strong as its weakest step. Direction matters too. Proving "if n is even, then n2 is even" does not prove "if n2 is even, then n is even". That second claim, the converse, happens to be true, but it needs its own argument.
One counterexample defeats a universal claim, while a million examples can't establish one. Euler's formula n2+n+41 gives primes for n=0,1,…,39, forty cases in a row, yet n=40 gives 1681=412. A derivation, such as working out the quadratic formula, can serve as a proof, but only if every step holds in every case, for example never dividing by something that could be zero.
Think of it like this. A proof is like a contract with no loopholes: every clause follows from agreed definitions, so no future case can wriggle out of it. Testing examples is like asking a few people whether they found a loophole yet. Unlike contracts, a correct proof is never overturned by a later case; only a flawed step can break it.
Core idea. A proof is a rigorous chain of valid inferences from accepted axioms to a guaranteed conclusion.