What does Thresholds and Budgets for Ambient Adjutants conclude, in numbers, about “Risk-Tiered Confirmation: Fatigue and the Undo Window”?

Answer: Holding for 5 seconds catches almost nothing (0.014) and has a negative net value. Ten seconds catch 0.062, 30 seconds 0.290, 60 seconds 0.493, 120 seconds 0.648 and 300 seconds 0.734, but by five minutes the cost of delay exceeds the gain and the net value turns negative. The best hold is 74 seconds, catching 0.548 for a net value of 9.04 per thousand sends. The 60-second undo window used for reversible actions in program p62, with a catch rate of 0.49, comes from this calculation.

Scope of the numbers: Parameters are illustrative and data synthetic; all calculations are generated by the book's 23 standard-library programs. The book is not individual business advice or legal advice; capture is limited to devices and scenes the wearer controls, bystander content is not kept by default or is redacted at once, and no effect on hearing or the body is claimed.

中文版

Data table

The numbers below are printed verbatim by the chapter's Python program and can be reproduced by running it. standard-library Python onlyprinted values that match the answer: 0.014, 0.062, 0.290, 0.49, 0.493, 0.548

hold w sP(caught)net per 1,000 sends
50.014-0.08
100.0620.86
200.1833.49
300.2905.70
600.4938.79
1200.6487.44
3000.734-7.99
best hold 74 s, P(caught) 0.548, net 9.04 per 1,000 sends
60 s window catch rate used in p62: 0.49
Show the full program
# Chapter 6. p63: sizing the hold-and-undo window for an outbound send.
# A wrong send is noticed after T seconds: with probability 0.75 T is
# log-normal (median 40 s, sigma 1.0); otherwise it is not noticed
# within the hour. Holding every send for w seconds catches a wrong one
# with P(T <= w) and delays every send by w. Per send:
#   net(w) = e P(T <= w) - h w,  e = 0.03 wrong sends, h = 0.0001 per s
# (the error rate and the price of delay in caught-error units are
# illustrative). Undo is a fail-safe default in the sense of Saltzer and
# Schroeder (1975): the safe state is the one reached by doing nothing.
import math
from statistics import NormalDist

E, H = 0.03, 0.0001
Z = NormalDist()


def caught(w):
    return 0.75 * Z.cdf(math.log(w / 40.0) / 1.0)


if __name__ == "__main__":
    print("  hold w s   P(caught)   net per 1,000 sends")
    for w in (5, 10, 20, 30, 60, 120, 300):
        print("  {:>8}    {:.3f}       {:>7.2f}".format(
            w, caught(w), 1000.0 * (E * caught(w) - H * w)))
    best = max(range(1, 601), key=lambda w: E * caught(w) - H * w)
    print("best hold {} s, P(caught) {:.3f}, net {:.2f} per 1,000 sends"
          .format(best, caught(best), 1000.0 * (E * caught(best) - H * best)))
    print("60 s window catch rate used in p62: {:.2f}".format(caught(60)))

Single source and citation

Fan, Y. (2026). Ambient Multimodal Cognitive Adjutants: Low-Latency Audio-Visual Memory Streams, Proactive Intent Anticipation and Real-Time Decision Support (Version 1.0.0), Chapter 6: Risk-Tiered Confirmation: Fatigue and the Undo Window. 5D Library. DOI pending

DOI pending · monograph page goes live after publication · 5DL-B-409 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026409,
  author    = {Fan, Yafu},
  title     = {Ambient Multimodal Cognitive Adjutants: Low-Latency Audio-Visual Memory Streams, Proactive Intent Anticipation and Real-Time Decision Support},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

Other chapter answers in this book