What does Thresholds and Budgets for Ambient Adjutants conclude, in numbers, about “End-to-End Latency Budgets: Percentiles, Not Slogans”?
Answer: Section A is the itemised bill for the baseline architecture. The largest item is generation of the first token, with a median of 379 milliseconds, a 95th percentile of 797 and a 99th percentile of 1,073; it also has the longest tail. Next comes the endpoint wait, with a median of 260 milliseconds and a 95th percentile of 328, the floor described in Chapter 2, which barely moves. Recognition has a median of 160 milliseconds. Retrieval has a median of only 70, but its 95th percentile jumps to 219, the widest spread of any stage. The total has a median of 1,083 milliseconds, a 95th percentile of 1,540 and a 99th percentile of 1,820.
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.
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: 1073, 1083, 1540, 160, 1820, 219
| A. baseline stages, ms | p50 | p95 | p99 |
| frame + features | 20 | 20 | 20 |
| endpoint wait | 260 | 328 | 404 |
| speech recognition | 160 | 284 | 360 |
| retrieval | 70 | 219 | 353 |
| answer first token | 379 | 797 | 1,073 |
| speech synthesis | 110 | 179 | 218 |
| delivery | 35 | 53 | 63 |
| total | 1,083 | 1,540 | 1,820 |
| B. variants, total ms p50 p95 p99 <=1s <=1.5s | |||
| baseline 1,083 1,540 1,820 0.323 0.938 | |||
| streaming 966 1,411 1,694 0.569 0.970 | |||
| stream+card 846 1,283 1,562 0.758 0.986 | |||
| stream+card+h120 726 1,163 1,442 0.873 0.993 | |||
Show the full program
# Chapter 3. p31: end-to-end latency budget by Monte Carlo.
# Stage time = fixed + median * exp(sigma * Z), Z standard normal, so
# each stage is a shifted log-normal. Stages run in sequence from the
# moment the user stops speaking to the first cue the user perceives.
# The endpoint wait carries the 240 ms hangover chosen in Chapter 2.
# All medians and sigmas (ms) are illustrative, not measurements of
# any product. 20,000 draws, fixed seed.
import math
import random
N = 20000
BASE = [ # name, fixed ms, median ms, sigma
("frame + features", 20.0, 0.0, 0.0),
("endpoint wait", 240.0, 20.0, 0.9),
("speech recognition", 0.0, 160.0, 0.35),
("retrieval", 0.0, 70.0, 0.7),
("answer first token", 0.0, 380.0, 0.45),
("speech synthesis", 0.0, 110.0, 0.3),
("delivery", 0.0, 35.0, 0.25),
]
def variant(name):
st = [list(s) for s in BASE]
if name in ("streaming", "stream+card", "stream+card+h120"):
st[2] = ["speech recognition", 0.0, 50.0, 0.4]
if name in ("stream+card", "stream+card+h120"):
st[5] = ["speech synthesis", 0.0, 0.0, 0.0]
st[6] = ["delivery", 0.0, 30.0, 0.25]
if name == "stream+card+h120":
st[1] = ["endpoint wait", 120.0, 20.0, 0.9]
return st
def pct(xs, q):
s = sorted(xs)
return s[int(round(q * (len(s) - 1)))]
def simulate(stages, rng):
totals, per = [], [[] for _ in stages]
for _ in range(N):
parts = []
for j, (_, fx, med, sg) in enumerate(stages):
v = fx + (med * math.exp(sg * rng.gauss(0.0, 1.0))
if med > 0 else 0.0)
parts.append(v)
per[j].append(v)
totals.append(math.fsum(parts))
return totals, per
if __name__ == "__main__":
rng = random.Random(40931)
totals, per = simulate(variant("baseline"), rng)
print("A. baseline stages, ms p50 p95 p99")
for (name, _, _, _), xs in zip(BASE, per):
print(" {:<24} {:>7,.0f} {:>7,.0f} {:>7,.0f}".format(
name, pct(xs, 0.5), pct(xs, 0.95), pct(xs, 0.99)))
print(" {:<24} {:>7,.0f} {:>7,.0f} {:>7,.0f}".format(
"total", pct(totals, 0.5), pct(totals, 0.95), pct(totals, 0.99)))
print("B. variants, total ms p50 p95 p99 <=1s <=1.5s")
for name in ("baseline", "streaming", "stream+card",
"stream+card+h120"):
rng = random.Random(40931)
t, _ = simulate(variant(name), rng)
a = 0
b = 0
for v in t:
a += v <= 1000.0
b += v <= 1500.0
print(" {:<20} {:>7,.0f} {:>7,.0f} {:>7,.0f} {:.3f} {:.3f}"
.format(name, pct(t, 0.5), pct(t, 0.95), pct(t, 0.99),
a / N, b / N))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 3: End-to-End Latency Budgets: Percentiles, Not Slogans. 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
- Chapter 1 · The Hit-Rate Threshold: When Speaking Up Pays
- Chapter 2 · Voice Activity Detection: the Endpoint Wait as a Latency Floor
- Chapter 4 · Speculative Prefetch: Read-Only, Break-Even and Server Load
- Chapter 5 · The Rolling Memory Window: Forgetting Curves, Bystander Exposure and One-Tap Deletion
- Chapter 6 · Risk-Tiered Confirmation: Fatigue and the Undo Window
- Chapter 7 · Seeing the First Sign and the Palaces of Memory: an East-West Source Critique
- Chapter 8 · The Adjutant Event Ledger: Field Tools and Checker Specifications