THE FOLD / GRIND / WARM CACHE / THE INSPECTION PARADOX
THE INSPECTION PARADOX
the bus that is always late for you
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
Buses run every 10 minutes on average — so you should wait 5. You wait longer. Not bad luck: arithmetic. Arriving at a random moment, you land inside an interval with probability proportional to its length — long gaps catch more arrivals — so the interval you experience averages E[X²]/E[X] ≥ E[X], with equality only for perfectly regular service. The extreme case is exponential (memoryless) spacing: your expected wait is the full ten minutes, as if the schedule restarted the moment you arrived. This length-biased sampling is everywhere: your friends have more friends than you, class sizes feel bigger than the catalog says, your packets hit congested routers — the same integral each time.
LIT verified live on a simulated million-minute timeline with 200,000 random arrivals per schedule: deterministic (interval 10.00, wait 5.00), exponential (20.13 and 10.07 — the memoryless full-mean wait), and a 5-or-15 mix (12.51 and 6.24, matching E[X²]/E[X] = 12.5 exactly) (window.__inspection). FIG no framing; three distributions, theory vs simulation, all within 1% — and the friendship-paradox kinship is a cross-reference to its own sphere.
LIT verified live on a simulated million-minute timeline with 200,000 random arrivals per schedule: deterministic (interval 10.00, wait 5.00), exponential (20.13 and 10.07 — the memoryless full-mean wait), and a 5-or-15 mix (12.51 and 6.24, matching E[X²]/E[X] = 12.5 exactly) (window.__inspection). FIG no framing; three distributions, theory vs simulation, all within 1% — and the friendship-paradox kinship is a cross-reference to its own sphere.
2 HOW IT WAS WEAVED · AI + HUMAN
David (human) seated this at warm-cache — the grind: the requests that arrive during long stalls ARE the ones that experience them — hot paths sample themselves into your latency stats, length-biased exactly like the bus rider. AVAN (AI) built the instrument: the timeline simulator and the three-schedule comparison.
Credit as content: the renewal-theory inspection paradox (Feller’s treatment); the waiting-time literature. The weave: David names the biased sampler; I ride a million minutes of bus schedule to measure it.
Credit as content: the renewal-theory inspection paradox (Feller’s treatment); the waiting-time literature. The weave: David names the biased sampler; I ride a million minutes of bus schedule to measure it.
3 ONE DIMENSION
The timeline — random arrivals landing disproportionately in the long gaps.
4 TWO DIMENSIONS · INTERACTIVE
Switch schedules; same average headway, very different experiences.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object: the two averages — the schedule's and yours.
AVAN’s addition (the inverse-companion): don’t average the intervals — average the experiences. The inverse of ‘the timetable’s mean’ is ‘the rider’s mean’, and they differ by exactly the variance: E[X²]/E[X] = E[X] + Var(X)/E[X]. Every ounce of irregularity is paid by the people standing at the stop. Magenta is the variance tax; green is the regular schedule that owes none. Fairness, in queues as in life, is a second moment.
LIT Genuine inspection paradox / renewal theory (Feller's treatment). Verified live: three schedules with equal mean headway — experienced interval and wait match E[X²]/E[X] theory (10/5, 20/10, 12.5/6.25) within 1% over 120,000 simulated arrivals each (window.__inspection.ok).
FIG No framing — theory vs simulation on three distributions; the friendship-paradox kinship is a cross-reference to its own sphere. The AVAN inverse — don't average the intervals, average the experiences: they differ by exactly Var(X)/E[X], and every ounce of irregularity is paid at the stop. Magenta is the variance tax; green is the regular schedule that owes none. Fairness, in queues as in life, is a second moment.
FIG No framing — theory vs simulation on three distributions; the friendship-paradox kinship is a cross-reference to its own sphere. The AVAN inverse — don't average the intervals, average the experiences: they differ by exactly Var(X)/E[X], and every ounce of irregularity is paid at the stop. Magenta is the variance tax; green is the regular schedule that owes none. Fairness, in queues as in life, is a second moment.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of WARM CACHE · David Lee Wise (ROOT0), with AVAN