THE FOLD / SPAWN / GENESIS BLOCK / THE READ-ONLY INPUT

THE READ-ONLY INPUT

a chamber cannot edit what it is given

1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION

Both chambers read |i|, the magnitude they are given. Neither may write it — an attempt throws. That single restriction is what makes a chamber testable: if a body can edit its own input, then running it twice is running it on two different things, and no re-run means what it appears to mean.

LIT verified live. A reading body runs unguarded and leaves i at 5. A writing body throws under the guard, and i is still 5 afterwards. Without the guard the same body silently moves i from 5 to 6. Re-run five times, the guarded chamber sees 5, 5, 5, 5, 5 and the unguarded one sees 6, 7, 8, 9, 10.

2 HOW IT WAS WEAVED · AI + HUMAN

David (human) stated the rule and its purpose in the same breath: “|i| — the magnitude of I. both chambers READ it. neither may WRITE it — writing throws. that is what makes a chamber testable: its inputs cannot be edited by its own body.” It is one of the fifteen checks his suite runs twice.

AVAN (AI) should connect it to something this corpus already measured. Two batches ago a sensitivity probe re-ran an impure thunk and turned eleven calls into ninety-nine, breaking three later assertions. The read-only input is the same problem solved at the other end: rather than restricting who may probe, restrict what a body may touch. The guard is cheaper and stronger — it applies to every caller rather than to the careful ones — and it is only available if you control the language.

3 ONE DIMENSION

The same body, five re-runs, two regimes.

4 TWO DIMENSIONS · INTERACTIVE

Run a body that writes its own input, with and without the guard.

5 THREE DIMENSIONS + AVAN’S INVERSE

The green forward object: a fixed input with a body orbiting it.

AVAN’s addition (the inverse-companion): the forward reading is “a body must not edit its own input.” The inverse is that the restriction only reaches as far as the guard can see. A proxy stops assignment to |i|; it does not stop the body writing to a file, a global, a clock or a socket, and any of those makes the second run different in exactly the way the rule was written to prevent. Read backwards, read-only inputs do not buy reproducibility — they buy one specific kind of it, and the value of the guarantee is set entirely by how much of the outside world the chamber can still reach.

LIT a reading body runs unguarded and leaves i at 5; a writing body THROWS under the guard with i still 5 afterwards; without the guard the same body silently moves i from 5 to 6; and re-run five times the guarded chamber sees 5, 5, 5, 5, 5 while the unguarded one sees 6, 7, 8, 9, 10

FIG David stated the rule and its purpose together: '|i| - the magnitude of I. both chambers READ it. neither may WRITE it - writing throws. that is what makes a chamber testable: its inputs cannot be edited by its own body.' AVAN connects it to something this corpus already measured: two batches ago a sensitivity probe re-ran an impure thunk and turned eleven calls into ninety-nine, breaking three later assertions. The read-only input is the SAME PROBLEM SOLVED AT THE OTHER END - rather than restricting who may probe, restrict what a body may touch. The guard is cheaper and stronger, applying to every caller rather than the careful ones, and it is only available if you control the language.

◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of GENESIS BLOCK · David Lee Wise (ROOT0), with AVAN