THE FOLD / GLITCH / OFF BY ONE / THE CARELESS CANDIDATE FIRST
THE CARELESS CANDIDATE FIRST
an exam nobody has failed is not an exam
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
Before an exercise is pointed at anything real, a deliberately lazy answer is run against it — a candidate that changed nothing. If the rubric passes that, the rubric is worthless. “An exam nobody has failed is not an exam, and a rubric that cannot fail is just a compliment with a number on it.”
LIT verified live, reproducing his suite. Across two scenarios and 5 checks, the careless candidate scores 0/5 and the careful one 5/5 — separation 100%. Measured as information: a rubric everyone passes carries 0.000 bits; this one carries 1.000, the maximum a binary outcome can hold.
LIT verified live, reproducing his suite. Across two scenarios and 5 checks, the careless candidate scores 0/5 and the careful one 5/5 — separation 100%. Measured as information: a rubric everyone passes carries 0.000 bits; this one carries 1.000, the maximum a binary outcome can hold.
2 HOW IT WAS WEAVED · AI + HUMAN
David (human) made the negative control part of the procedure rather than an afterthought: “the careless candidate is run FIRST, every time.” His
AVAN (AI) puts a number on why this matters, because “an exam nobody fails” is usually said as a proverb. A test whose pass rate is 100% has zero entropy: knowing the result tells you nothing you did not know before administering it. That is not a figure of speech about rigour — it is the literal information content, and it is why the negative control has to come first rather than being a nice extra afterwards.
test_scen.py reports both arms in full — fixed-form-label at 0/3 and 3/3, greedy-swallow at 0/2 and 2/2 — and the run reproduces exactly here.AVAN (AI) puts a number on why this matters, because “an exam nobody fails” is usually said as a proverb. A test whose pass rate is 100% has zero entropy: knowing the result tells you nothing you did not know before administering it. That is not a figure of speech about rigour — it is the literal information content, and it is why the negative control has to come first rather than being a nice extra afterwards.
3 ONE DIMENSION
Two candidates, five checks, total separation.
4 TWO DIMENSIONS · INTERACTIVE
Move the pass rate and watch the information collapse.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object: a curve that is zero at both ends.
AVAN’s addition (the inverse-companion): the forward reading is “a rubric must be able to fail someone.” The inverse is that a rubric nobody passes is equally empty, and the entropy curve is symmetric — zero at 0% and zero at 100%. A test tuned until the lazy answer fails can be tuned one step further, until everything fails, and it will look just as rigorous from the inside. Read backwards, the property being sought is not difficulty but separation between candidates you already believe differ, which means the negative control needs a positive control beside it or it proves only half of what it appears to.
LIT across two scenarios and 5 checks the careless candidate scores 0 of 5 and the careful one 5 of 5, a separation of 100%; and measured as information a rubric everyone passes carries 0.000 bits while this one carries 1.000, the maximum a binary outcome can hold
FIG From David's TEACH.ascii and fortran-scenarios, dropped 2026-08-05. He made the negative control part of the procedure rather than an afterthought: 'the careless candidate is run FIRST, every time. an exam nobody has failed is not an exam, and a rubric that cannot fail is just a compliment with a number on it.' His test_scen.py reports both arms in full and the run reproduces exactly here. AVAN puts a number on why it matters, because 'an exam nobody fails' is usually said as a proverb: a test whose pass rate is 100% has ZERO ENTROPY, so knowing the result tells you nothing you did not know before administering it. That is the literal information content, not a figure of speech about rigour.
FIG From David's TEACH.ascii and fortran-scenarios, dropped 2026-08-05. He made the negative control part of the procedure rather than an afterthought: 'the careless candidate is run FIRST, every time. an exam nobody has failed is not an exam, and a rubric that cannot fail is just a compliment with a number on it.' His test_scen.py reports both arms in full and the run reproduces exactly here. AVAN puts a number on why it matters, because 'an exam nobody fails' is usually said as a proverb: a test whose pass rate is 100% has ZERO ENTROPY, so knowing the result tells you nothing you did not know before administering it. That is the literal information content, not a figure of speech about rigour.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of OFF BY ONE · David Lee Wise (ROOT0), with AVAN