THE FOLD / CHEAT / THE ROOT KIT / THE WEIGHTED ARM
THE WEIGHTED ARM
one measurement, two headlines
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
Two arms, measured once and reported twice. Raw, arm A is 16,803 and arm B is 29,416 — a ratio of 1.751. Weighted, A is unchanged and B becomes 72,686 — a ratio of 4.326. Both figures are correct and they answer different questions, and the entire distance between them is one number: arm B’s mean weight. Reporting only one of the two would have been a decision about which question mattered, taken silently.
LIT verified live: the raw ratio is 1.751 and the weighted ratio 4.326, both matching his receipts exactly; arm A is untouched by weighting while arm B is multiplied by 2.471, and that factor is the amplification between the two reported ratios exactly; the weight model in the same file (H=1, S=3, O=5) gives a first-to-last ratio of 0.2 = 1/5 and not 1/3, as his receipts state; and B’s recovered mean weight of 2.471 lies inside the model’s own range of 1 to 5.
LIT verified live: the raw ratio is 1.751 and the weighted ratio 4.326, both matching his receipts exactly; arm A is untouched by weighting while arm B is multiplied by 2.471, and that factor is the amplification between the two reported ratios exactly; the weight model in the same file (H=1, S=3, O=5) gives a first-to-last ratio of 0.2 = 1/5 and not 1/3, as his receipts state; and B’s recovered mean weight of 2.471 lies inside the model’s own range of 1 to 5.
2 HOW IT WAS WEAVED · AI + HUMAN
David (human) printed both ratios in receipts-python.txt and a line noting that the repository’s own weight model implies 1/5 rather than 1/3 — an internal consistency check where one of the subject’s files confirms another. Seated at THE ROOT KIT, because a weighting applied downstream can change a headline number without touching a single measurement.
AVAN (AI) recovered the mean weight rather than being told it. Arm A being identical raw and weighted pins it — every item in A carries weight 1 — so B’s multiplier falls straight out of 72,686 ÷ 29,416 = 2.471, and that is exactly the ratio between 4.326 and 1.751. It also lands inside the declared range of 1 to 5, so the two halves of the file agree with each other, which is the kind of check worth doing precisely because it usually passes and costs nothing. What is not established here is which ratio is the right one to quote; that depends on what the arms are for, and nothing in the arithmetic decides it.
AVAN (AI) recovered the mean weight rather than being told it. Arm A being identical raw and weighted pins it — every item in A carries weight 1 — so B’s multiplier falls straight out of 72,686 ÷ 29,416 = 2.471, and that is exactly the ratio between 4.326 and 1.751. It also lands inside the declared range of 1 to 5, so the two halves of the file agree with each other, which is the kind of check worth doing precisely because it usually passes and costs nothing. What is not established here is which ratio is the right one to quote; that depends on what the arms are for, and nothing in the arithmetic decides it.
3 ONE DIMENSION
The same pair of arms, before and after weighting.
4 TWO DIMENSIONS · INTERACTIVE
Turn the weighting on and off and watch a headline ratio move without any data changing.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object: one measurement, two heights, depending on a multiplier applied later.
AVAN’s addition (the inverse-companion): the forward reading is “report both ratios.” The inverse is that a weighting is a claim, and it is the least visible kind. The raw counts can be audited; the weights are a judgement about what counts for how much, applied after the measuring is done, and they moved this headline by a factor of 2.5 without a single observation changing. Read backwards, the reason to print the raw figure beside the weighted one is not redundancy — it is that their ratio is the only place the weighting becomes a number anyone can argue with.
LIT the raw ratio is 1.751 and the weighted ratio 4.326, both matching his receipts exactly; arm A is untouched by weighting while arm B is multiplied by 2.471, and that factor IS the amplification between the two reported ratios exactly; the weight model in the same file (H=1, S=3, O=5) gives a first-to-last ratio of 0.2 = 1/5 and not 1/3, as his receipts state; and B's recovered mean weight of 2.471 lies inside the model's own range of 1 to 5
FIG The mean weight was RECOVERED, not given. Arm A being identical raw and weighted pins every item in A at weight 1, so B's multiplier falls straight out of 72,686 / 29,416 = 2.471, exactly the ratio between 4.326 and 1.751. It also lands inside the declared range, so the two halves of the file agree. What is NOT established here is which ratio is the right one to quote — that depends on what the arms are for, and nothing in the arithmetic decides it.
FIG The mean weight was RECOVERED, not given. Arm A being identical raw and weighted pins every item in A at weight 1, so B's multiplier falls straight out of 72,686 / 29,416 = 2.471, exactly the ratio between 4.326 and 1.751. It also lands inside the declared range, so the two halves of the file agree. What is NOT established here is which ratio is the right one to quote — that depends on what the arms are for, and nothing in the arithmetic decides it.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of THE ROOT KIT · David Lee Wise (ROOT0), with AVAN