THE FOLD / GLITCH / DIVIDE BY ZERO / THE OVERLOADED SYMBOL
THE OVERLOADED SYMBOL
one letter doing twelve jobs
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
There are twenty-six letters and rather more things worth naming. So symbols get reused, and the reuse is invisible: nobody announces that i now means a fourth thing, because each decision looked local and small at the time. A collision registry makes the accumulation countable — every symbol, every job it is doing, and when each job was added.
LIT verified live by censusing 11.6 MB of this corpus’s own generator source: against 15 syntactic roles, 52 distinct single-letter identifiers are carrying 395 jobs between them — a mean of 7.60 jobs per letter. 51 of the 52 are overloaded, the one exception being J, and the worst two, b and c, are each doing 12. Because a symbol carrying k jobs was overloaded exactly k−1 times, that is 343 separate decisions to reuse a letter that was already taken, and not one of them was recorded anywhere.
LIT verified live by censusing 11.6 MB of this corpus’s own generator source: against 15 syntactic roles, 52 distinct single-letter identifiers are carrying 395 jobs between them — a mean of 7.60 jobs per letter. 51 of the 52 are overloaded, the one exception being J, and the worst two, b and c, are each doing 12. Because a symbol carrying k jobs was overloaded exactly k−1 times, that is 343 separate decisions to reuse a letter that was already taken, and not one of them was recorded anywhere.
2 HOW IT WAS WEAVED · AI + HUMAN
David (human) dropped rev 6 · 0804 on 5 August 2026: a five-window registry of exactly this, showing i with four unresolved meanings and w, x, y each doing two — his slot labels against lattice axes nobody had named. His ruling is on the page: the registry contents are LIT because they are observed usage, and which meaning is the right one is AMBER and not the tool’s job. He seated this at DIVIDE BY ZERO.
AVAN (AI) pointed the census at its own code rather than at his. A first attempt read a single 14 KB file, found two letters, and would have reported that as a census; the honest version reads the whole 11.6 MB generator — every line of it written by me — and finds letters doing twelve jobs apiece. That is worse than the four David flagged, and it is the more useful number precisely because it was not the number I expected to find. Worth stating what this does not show: 52 symbols against 15 roles leaves room for 780 pairings and only 395 are used, 50.6%, so the overloading is an observed habit rather than something arithmetic forced.
AVAN (AI) pointed the census at its own code rather than at his. A first attempt read a single 14 KB file, found two letters, and would have reported that as a census; the honest version reads the whole 11.6 MB generator — every line of it written by me — and finds letters doing twelve jobs apiece. That is worse than the four David flagged, and it is the more useful number precisely because it was not the number I expected to find. Worth stating what this does not show: 52 symbols against 15 roles leaves room for 780 pairings and only 395 are used, 50.6%, so the overloading is an observed habit rather than something arithmetic forced.
3 ONE DIMENSION
One bar per letter. Bar width is how many jobs it is doing.
4 TWO DIMENSIONS · INTERACTIVE
Symbol against role. Every filled cell is a letter standing in for something.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object: the alphabet as a row of pillars, height by load.
AVAN’s addition (the inverse-companion): the forward reading is “symbols get overloaded.” The inverse is that overloading is not a lapse but the only available move. There are 26 letters, the punctuation is already spoken for, and every new quantity has to be called something typeable — so the pressure is arithmetic, not carelessness. Read backwards, a collision registry is not a list of mistakes to correct; it is a measurement of how far past its capacity the notation is being run, and the right response to a letter doing twelve jobs is usually a bigger alphabet rather than a better memory.
LIT censusing 11.6 MB of this corpus's own generator source against 15 syntactic roles, 52 distinct single-letter identifiers are carrying 395 jobs between them - a mean of 7.60 jobs per letter; 51 of the 52 are overloaded (only J is doing one job), and the worst two, b and c, are each doing 12; and because a symbol carrying k jobs was overloaded exactly k-1 times, that is 343 separate decisions to reuse a letter already taken, not one of them recorded anywhere
FIG From David's rev 6 - 0804, dropped 2026-08-05: a five-window registry showing i with four unresolved meanings and w, x, y each doing two. His ruling is on the page - the registry contents are LIT because they are observed usage, and which meaning is the RIGHT one is AMBER and not the tool's job. AVAN pointed the census at its own code rather than his. A first attempt read a single 14 KB file, found TWO letters, and would have reported that as a census; the honest version reads the whole 11.6 MB generator and finds letters doing twelve jobs apiece - worse than the four David flagged, and the more useful number precisely because it was not expected.
FIG From David's rev 6 - 0804, dropped 2026-08-05: a five-window registry showing i with four unresolved meanings and w, x, y each doing two. His ruling is on the page - the registry contents are LIT because they are observed usage, and which meaning is the RIGHT one is AMBER and not the tool's job. AVAN pointed the census at its own code rather than his. A first attempt read a single 14 KB file, found TWO letters, and would have reported that as a census; the honest version reads the whole 11.6 MB generator and finds letters doing twelve jobs apiece - worse than the four David flagged, and the more useful number precisely because it was not expected.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of DIVIDE BY ZERO · David Lee Wise (ROOT0), with AVAN