Why one of the two means can never overtake the other

0:00Take two numbers above nothing — four and sixteen. There are two natural numbers to put between them, and they are not the same number. Add the two and halve: ten. That is the half-sum, the average you have been using since primary school, and this is where it comes in from — nothing here builds it. Now multiply the two and take the square root. Sixty-four, and the root of sixty-four is eight.

0:29That is the middle that makes four, eight, sixteen a progression, multiplying by two at each step. Ten and eight. Two middles for one pair, and the adding one came out the larger. Try another pair and it happens again. And again. The question is whether it always happens, and why. Notice what that question is not. It is not whether it happens for these numbers. It is whether it happens for every pair of numbers above nothing — and no amount of trying settles a claim of that shape.

1:07That is worth a demonstration, because it sounds like fussiness and it is not. The checker behind this video builds a second expression on purpose. On a grid of nine sample points it agrees with the true one exactly, every time — nine out of nine. If those nine were your evidence, you would conclude that the expression is never below nothing. Step off the grid and it takes the value minus twenty.

1:37Nine agreeing examples told you nothing at all about a sign. So we are going to prove it instead. It takes three lines. Line one. Subtract one middle from the other. The half-sum is a plus b, over two. The geometric middle is the square root of a times b. Take the second from the first and put the whole thing over one denominator. The result is a plus b, minus twice the square root of a b, all over two.

2:10Stop there and look at the numerator. The entire argument is about to happen inside it, and nothing after this line is difficult. Line two is a recognition, and it is the step people accept and cannot reproduce, because it usually arrives already recognised. So do it in the direction you would need it. Write s for the square root of a, and t for the square root of b.

2:40Then a is s squared, b is t squared, and the square root of a b is s times t. Now take s minus t, and square it. s squared, minus two s t, plus t squared. Set that beside the numerator: a, minus twice the root of a b, plus b. They are the same expression. Not similar — the same, term for term. The checker compares them coefficient by coefficient, so that is a measurement rather than a resemblance.

3:14Line three, and it is a single fact. A real number, squared, is never below nothing. So the numerator is never below nothing. The denominator is two. The whole difference is never below nothing. The half-sum minus the geometric middle is at least nothing. Which is to say the half-sum is at least the geometric middle. Every pair of numbers above nothing, settled at once, and there is not one example anywhere in the argument.

3:47That is what a proof buys you that a hundred pairs cannot. Now the half that usually gets skipped. At least includes equal to, so it is worth asking when the equal case actually happens. A square is nothing exactly when the thing being squared is nothing. So the difference is nothing exactly when s minus t is nothing — when the two square roots agree. And for numbers above nothing, the roots agreeing means the numbers agree.

4:19Nine and nine. The half-sum is nine. The geometric middle is nine. That is the only way the two ever meet. For any other pair the half-sum is strictly the larger. The word above nothing, in all of this, is not decoration. Take minus one and minus four. Their half-sum is minus two and a half. Their product is four, and the square root of four is two. So here the half-sum is the smaller of the two. The comparison runs backwards.

4:54That is not a hole in the proof, because the proof was about numbers above nothing. It is a demonstration of what the condition is holding back. And the condition has a second job: it is what makes the product positive, so that s and t — the two square roots the argument is built on — are real numbers at all. The checker refuses to hand out a geometric middle for any pair that is not above nothing, so that refusal is structural rather than a note.

5:29Put all four numbers on a line and you can see the shape of the result. Four and sixteen at the ends. The half-sum, ten, sits exactly between them: six from each end. The geometric middle, eight, sits four above the left end and eight below the right. It leans towards the smaller number, and it always will — leaning towards the smaller and being the smaller of the two middles are the same fact.

5:59Now slide the two ends towards each other. The gap between the middles closes, and when the ends meet, the middles meet with them. Now run the whole thing backwards, because these two middles carry more than they look like they carry. Double the half-sum and you have the total of the pair. Square the geometric middle and you have their product. And a total and a product are between them enough to pin two numbers down.

6:29So a pair of numbers and its pair of middles hold exactly the same information. Give me the middles and I can give you back the numbers. Here is that done. Two numbers above nothing whose half-sum is ten and whose geometric middle is eight. The total is twenty. The product is sixty-four. Now one identity: the square of the gap between two numbers is the square of their total, less four times their product.

7:01Four hundred, minus two hundred and fifty-six, is one hundred and forty-four. So the gap is twelve. Total twenty, gap twelve — the numbers are four and sixteen. Check them. Half-sum ten. Product sixty-four, whose root is eight. Both middles come back. And that identity is not a separate thing to memorise. It is the expansion of a squared difference, rearranged. One thing about that answer deserves saying plainly. The square of the gap was one hundred and forty-four, and that has two roots: twelve and minus twelve.

7:43Twelve gives four and then sixteen. Minus twelve gives sixteen and then four. That is not two answers. It is one pair written twice, because nothing in the question said which number to call the first. The checker searched four hundred and fifteen candidate numbers and found exactly two ordered pairs — and exactly one pair. And now the reason the comparison is doing real work, rather than sitting there as a remark.

8:16Ask for two numbers above nothing whose half-sum is eight and whose geometric middle is ten. The same two values, swapped over. The total would be sixteen. The product would be one hundred. The square of the gap: two hundred and fifty-six, minus four hundred. Minus one hundred and forty-four. No real number squares to something below nothing, so there is no such pair. Not hard to find. There is nothing there to find.

8:48The search agrees — it returns nothing at all — and the closed form refuses outright rather than handing back a root it cannot take. So the comparison is a gate on which questions have answers. That is why its direction is the whole content. You will meet the same fact wearing other clothes. Show that the two numbers are the half-sum, plus or minus the square root of: the half-sum plus the geometric middle, times the half-sum minus the geometric middle.

9:25That square root is real only because the second bracket is not below nothing. The comparison, doing the work, quietly. Or: two numbers whose total is six times their geometric middle. That says the half-sum is three times the geometric middle, and the pair comes out as three plus two root two, and three minus two root two. Multiply those together and you get exactly one. Add them and halve and you get exactly three. Three times one.

9:59Or: the two middles standing in a given ratio, m to n. The answer needs the square root of m squared minus n squared — real only when m is at least n. The comparison again, this time deciding whether the question has an answer at all. Three lines, one square, and one fact about squares. That is the whole of it, and it is why you never have to check it on a pair again.