THE FOLD / CO-OP / THE BROADCAST / THE HADAMARD
THE HADAMARD
a ±1 matrix with every row orthogonal — H·Hᵀ = nI
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
A Hadamard matrix is a square grid filled with only +1 and −1 whose rows are all mutually orthogonal: any two different rows agree in exactly half their entries and disagree in the other half, so their dot product is zero. Compactly, H·Hᵀ = nI.
Sylvester’s doubling builds one at every power of two: start with [1], then repeatedly tile four copies in a 2×2 block with the bottom-right negated. The rows are the Walsh / Hadamard codes — perfectly non-interfering signals that let many transmitters share one channel at once (the maths behind CDMA). They also form an error-correcting code: the [32,6] Hadamard code flew aboard Mariner 9 to beam photographs back from Mars through heavy noise. Orthogonality is the whole trick — mix the coded streams together and each can be pulled back out cleanly.
LIT verified live: the Sylvester matrices up to 32×32 have only ±1 entries and satisfy H·Hᵀ = nI (every distinct row-pair orthogonal), and H is its own inverse up to the factor 1/n (window.__hadamard). FIG no framing; the orthogonality and self-inverse are exact.
Sylvester’s doubling builds one at every power of two: start with [1], then repeatedly tile four copies in a 2×2 block with the bottom-right negated. The rows are the Walsh / Hadamard codes — perfectly non-interfering signals that let many transmitters share one channel at once (the maths behind CDMA). They also form an error-correcting code: the [32,6] Hadamard code flew aboard Mariner 9 to beam photographs back from Mars through heavy noise. Orthogonality is the whole trick — mix the coded streams together and each can be pulled back out cleanly.
LIT verified live: the Sylvester matrices up to 32×32 have only ±1 entries and satisfy H·Hᵀ = nI (every distinct row-pair orthogonal), and H is its own inverse up to the factor 1/n (window.__hadamard). FIG no framing; the orthogonality and self-inverse are exact.
2 HOW IT WAS WEAVED · AI + HUMAN
David (human) seated this in THE BROADCAST — the co-op domain of many voices sharing one channel. A Hadamard matrix is exactly a broadcast trick: orthogonal codes that let everyone transmit at once and still be separated. AVAN (AI) built the instrument: the Walsh waveform, the checkerboard matrix, the encode/decode self-inverse.
The weave: David names the seat (many share one channel); I make the rows come out orthogonal and show a mixed signal separating cleanly — the waveform in 1D, the matrix in 2D, the self-inverse transform in 3D. The sphere is the seam. Credit: Jacques Hadamard (1893); Sylvester’s construction (1867); Walsh functions (1923); flown on Mariner 9 (1971).
The weave: David names the seat (many share one channel); I make the rows come out orthogonal and show a mixed signal separating cleanly — the waveform in 1D, the matrix in 2D, the self-inverse transform in 3D. The sphere is the seam. Credit: Jacques Hadamard (1893); Sylvester’s construction (1867); Walsh functions (1923); flown on Mariner 9 (1971).
3 ONE DIMENSION
Two Walsh rows as ±1 waveforms. Multiply them entry by entry and the pluses and minuses cancel exactly — the running sum returns to zero. That vanishing dot product is what “orthogonal” means, drawn on a line.
4 TWO DIMENSIONS · INTERACTIVE
The Hadamard matrix as a tile grid — white +1, black −1 — with its self-similar fractal pattern. Pick two rows and read their dot product: 0 for any two different rows, n for a row with itself. Grow the size and the orthogonality holds at every scale.
5 THREE DIMENSIONS + AVAN’S INVERSE
Several Walsh-coded streams summed into one noisy channel — the green forward step: many messages mixed together into a single broadcast.
AVAN’s addition (the inverse-companion): the magenta step pulls one stream back out — and it uses the very same matrix. Because H·H = nI, multiplying the mixed signal by a Walsh row cancels every other stream to zero and leaves just that one, scaled by n. The decode is the encode run again: H is, up to the factor 1/n, its own inverse. So the forward ‘mix everyone together’ and the inverse ‘separate one out’ are not two machines but one machine used twice — orthogonality is exactly the property that makes a transform undo itself. Green mixes the voices into a single channel; magenta applies the same Hadamard step and recovers a single voice untouched. The inverse of broadcasting is listening, and here they are the identical operation.
LIT Genuine Hadamard matrix (Jacques Hadamard 1893; Sylvester's construction 1867; Walsh functions 1923; the [32,6] Hadamard code flown on Mariner 9, 1971). Verified live: the Sylvester matrices up to 32x32 have only +-1 entries and satisfy H*H^T = nI (every distinct row-pair orthogonal, each row with itself = n), and H8*H8 = 8I so H is its own inverse up to 1/n (window.__hadamard.orthogonal && selfInverse). The orthogonality and self-inverse are exact.
FIG No framing: the +-1 entries, the mutual row-orthogonality (H*H^T = nI), and the self-inverse property are real and checked exhaustively over all row-pairs up to 32x32. The CDMA/Walsh-code and Mariner-9 error-correction uses are genuine documented applications; the encode-equals-decode inverse is demonstrated (mix streams, multiply by a row, recover one), not merely claimed.
FIG No framing: the +-1 entries, the mutual row-orthogonality (H*H^T = nI), and the self-inverse property are real and checked exhaustively over all row-pairs up to 32x32. The CDMA/Walsh-code and Mariner-9 error-correction uses are genuine documented applications; the encode-equals-decode inverse is demonstrated (mix streams, multiply by a row, recover one), not merely claimed.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of THE BROADCAST · David Lee Wise (ROOT0), with AVAN