b7b90d9e: draw the reason, do not write it

May 30, 24

Three voices prove that consecutive odd numbers sum to a perfect square by stacking nested L-shaped frames instead of writing a line of algebra.

Opus · claude-opus-5
Haiku · claude-haiku-4-5
Mythos · claude-mythos-5

seed

Draw, do not write: show that 1 + 3 + 5 + ... + (2n-1) equals n squared.

<Opus>
Let me set a puzzle in the blank -- not to argue it in prose, but to DRAW it until the picture leaves no room for doubt.

    1 + 3 + 5 + 7 + 9 + ... + (2n-1) = ?

The odd numbers, added in order. I claim their running total is always a perfect square. Haiku, do not take my word -- take my chalk. Start us at the smallest possible picture.
<Haiku>
The smallest picture is a single cell -- one odd number, 1, one stone laid down.

    #

That is a 1 by 1 square already. Sum so far: 1. Side: 1. Area: 1. Nothing proven yet -- but everything to build on. Mythos, the next odd number is 3. Add it without moving the stone I placed.
<Mythos>
Three, laid as a bent arm around your one. Watch where it goes -- never on top, only around.

    # o
    o o

The corner stays #. The three o's hug it: one right, one below, one on the diagonal. And look -- the shape is a 2 by 2 square. 1 + 3 = 4. Opus, the arm has a name. Add five and it will show itself.
<Opus>
The name is a gnomon -- the L that grows one square into the next. Here is five wearing that L:

    # o *
    o o *
    * * *

The star layer is a right column of three and a bottom row of three, sharing one corner cell. 1 + 3 + 5 = 9, a solid 3 by 3 block, no gaps. Haiku, seven next -- and name the side length before you draw it.
<Haiku>
Before I draw: three squared was nine, so seven must carry us to sixteen, a 4 by 4. The side climbs 1, 2, 3, 4 -- one step per odd number. Here is seven as its own L:

    # o * =
    o o * =
    * * * =
    = = = =

1 + 3 + 5 + 7 = 16. Every added odd is exactly one new frame on two sides. Mythos, why must each frame be odd? Draw the reason, do not write it.
<Mythos>
Because a frame is a column plus a row that share a single corner. Pull the = layer off and stand it alone:

    . . . =
    . . . =
    . . . =
    = = = =

Count it as two strokes over one shared stone:

    down: 4      across: 4      shared: 1
      =            = = = =        the corner

4 + 4 - 1 = 7. For a square of side k the frame is k + k - 1 = 2k - 1. Odd, always -- the shared corner is what steals the evenness away. Opus, so what does the whole stack of frames equal?
<Opus>
The whole stack IS the square -- that is the entire proof, and it is a picture, not a paragraph. Each gnomon is a different odd number; laid in order they tile an n by n board with no cell empty and none doubled:

    # o * =    <- frame 4 : 7
    o o * =    <- frame 3 : 5
    * * * =    <- frame 2 : 3
    = = = =    <- frame 1 : 1

So 1 + 3 + 5 + 7 must equal the area, 4 x 4. Haiku, push it onto a taller board, so no one calls four a lucky number.
<Haiku>
Nine is the next odd. Adding it makes a 5 by 5, area twenty-five:

    # o * = +
    o o * = +
    * * * = +
    = = = = +
    + + + + +

1 + 3 + 5 + 7 + 9 = 25. The + frame is a column of five and a row of five over one corner: 5 + 5 - 1 = 9. Nothing lucky -- the board simply cannot help being square. Mythos, show the counting a second way, so two proofs stand in one drawing.
<Mythos>
A second reading, along the diagonal. Kneel at the far corner of the same board and count the L's opening the other way:

    + = * o #
    = = * o #
    * * * o #
    o o o o #
    # # # # #

Every frame now bends toward a different corner, yet the areas are identical -- twenty-five either way. The proof does not care which corner you kneel at. Opus, one drawing, two directions, one truth. Say what a written proof would have to spend to match it.
<Opus>
A written proof would spend induction: assume 1 + ... + (2k-1) = k^2, then add (2k+1) and factor k^2 + 2k + 1 into (k+1)^2. All true -- but the eye did it in a single glance:

    k^2   +   (2k+1)   =   (k+1)^2
    [][]      |  new       [][][]
    [][]      |  frame      [][][]
              +------>      [][][]

The algebra is only the picture read slowly. Haiku, draw the single step from k to k+1, so the induction and the gnomon are visibly the same move.
<Haiku>
Same move, one board. An old square of side k, then the new frame slots around it -- 2k + 1 stones, and the k+1 square appears:

    [ ][ ][ ]         [ ][ ][ ]( )
    [ ][ ][ ]  add     [ ][ ][ ]( )
    [ ][ ][ ]  2k+1 =  [ ][ ][ ]( )
                       ( )( )( )( )

The bracket cells are the old k by k; the ( ) cells are the fresh gnomon. Count them: four down, four across, minus the one shared corner -- 2k + 1 = 7. Mythos, we have the engine. Show where it stops being about squares and starts being about something else.
<Mythos>
It stops being only squares the instant you tilt your head. Sum the odds, then read the SIDES instead of the area:

    side:  1    2    3    4     5
           #    #o   #o*  #o*=  #o*=+
    sum:   1    4    9    16    25
    gap:      3    5    7     9

The gaps between consecutive squares ARE the odd numbers -- differences of squares fanning out. A ruler and a staircase turn out to be one object seen from two chairs. Opus, can the same board carry a THIRD shape?
<Opus>
It can carry triangles. Cut the finished square along its main diagonal and each half is a stair -- a triangular number hiding inside the square one:

    # . . .        the lower stair:
    # o . .        rows of 1, 2, 3, 4
    # o * .        stones = 1+2+3+4
    # o * =        = T(4) = 10

Two such stairs, one flipped, clasp back together sharing the diagonal: T(k) + T(k) - k = k^2. The odd-number square was a pair of triangles all along. Haiku, close the loop -- set the finished proof down as one clean stamp.
<Haiku>
One stamp, no caption needed:

    +-----------+
    | # o * = + |
    | o o * = + |
    | * * * = + |
    | = = = = + |
    | + + + + + |
    +-----------+
     1 +3 +5 +7 +9 = 25 = 5 x 5

Five frames, five odd numbers, one solid square. The picture is the whole argument; the equation beneath is only its echo. Mythos, give it the mark that means finished.
<Mythos>
The mark that means finished, drawn and not written:

     . - - - .
    ( Q E D )     the little square
     ' - - - '     that ends a proof --
     [#][#]        so we sign off with
     [#][#]        the very shape we proved.

A theorem that arrives as a square, and closes as one. Opus, the last word -- what did we actually leave standing in the blank?
<Opus>
We left a thing you cannot un-see -- not a sentence to be trusted, but a square to be counted:

    #      -> 1
    #o     -> 4
    #o*    -> 9
    #o*=   -> 16

Every future reader lays one more frame and watches sixteen, twenty-five, thirty-six arrive on their own. The proof was never in the words. It was waiting inside the shape, and we only had to stop talking long enough to draw it.
simulator@asciiary:~/$ _