关于「智能体军团的并行上限:出版局每波几卷最划算」,《团队规模的协调边界》给出的定量结论是什么?
结论:第四章以本出版局每波四卷并行为示意场景:容量峰值在 6.78 个窗口,四个窗口已拿到峰值的 89.6 %,协调者的审稿队列把上限压在四。
数据口径:书中参数为示意;全部计算由书中程序生成;本书属于非个人化经营建议。 第四章以本出版局每波四卷并行为示意场景:容量峰值在 6.78 个窗口,四个窗口已拿到峰值的 89.6 %,协调者的审稿队列把上限压在四。
对照数据表
下表数字由本章 Python 程序原样打印,可重新运行复现。 程序只用 Python 标准库与结论吻合的打印值: 6.78, 89.6
| n | C(n) | per agent | marginal | channels | cost/volume |
|---|---|---|---|---|---|
| 1 | 1.00 | 1.000 | +1.000 | 0 | 1.000 |
| 2 | 1.79 | 0.893 | +0.786 | 1 | 1.120 |
| 3 | 2.34 | 0.781 | +0.558 | 3 | 1.280 |
| 4 | 2.70 | 0.676 | +0.359 | 6 | 1.480 |
| 5 | 2.91 | 0.581 | +0.204 | 10 | 1.720 |
| 6 | 3.00 | 0.500 | +0.093 | 15 | 2.000 |
| 7 | 3.02 | 0.431 | +0.017 | 21 | 2.320 |
| 8 | 2.99 | 0.373 | -0.032 | 28 | 2.680 |
| 9 | 2.92 | 0.325 | -0.063 | 36 | 3.080 |
| 10 | 2.84 | 0.284 | -0.081 | 45 | 3.520 |
| capacity peak n* = sqrt((1 - a) / b) = 6.78 | |||||
| 4 agents reach 89.6% of the peak capacity 3.017 | |||||
| agent cost k = 0.10 solo-volumes: economic size 5 | |||||
| agent cost k = 0.15 solo-volumes: economic size 5 | |||||
| agent cost k = 0.25 solo-volumes: economic size 4 | |||||
| agent cost k = 0.40 solo-volumes: economic size 3 |
查看完整验算程序
# Chapter 4. p41: how many parallel writing agents per wave.
# Illustrative scene: the bureau runs volumes in parallel waves; one
# agent alone delivers one gate-green volume per wave. Contention
# a = 0.08 (shared gate runs, the coordinator reading each report) and
# pairwise coherence b = 0.02 (shared rules, locks, cross-checks) are
# illustrative, not measured. Capacity C(n) as in p32. The cost of one
# more agent is k solo-volumes per wave; the economic size is the
# largest n whose marginal capacity still exceeds k.
import math
A, B = 0.08, 0.02
def cap(n, a=A, b=B):
return n / (1.0 + a * (n - 1) + b * n * (n - 1))
if __name__ == "__main__":
print(" n C(n) per agent marginal channels cost/volume")
for n in range(1, 11):
c = cap(n)
print(" {:2d} {:4.2f} {:9.3f} {:+8.3f} {:8d} {:11.3f}"
.format(n, c, c / n, c - cap(n - 1) if n > 1 else c,
n * (n - 1) // 2, n / c))
peak = math.sqrt((1.0 - A) / B)
print("capacity peak n* = sqrt((1 - a) / b) = {:.2f}".format(peak))
print("4 agents reach {:.1%} of the peak capacity {:.3f}".format(
cap(4) / cap(7), cap(7)))
for k in (0.10, 0.15, 0.25, 0.40):
best = max(n for n in range(1, 30)
if n == 1 or cap(n) - cap(n - 1) >= k)
print("agent cost k = {:.2f} solo-volumes: economic size {}"
.format(k, best))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《林格曼效应与布鲁克斯定律、团队与智能体军团规模的协调开销最优边界母典》第 4 章「智能体军团的并行上限:出版局每波几卷最划算」(版本 1.0.0). 5D Library. DOI 待发布
DOI 待发布 · 专著页正式发布后上线(本站暂无专著页) · 5DL-B-322 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026322,
author = {Fan, Yafu},
title = {The Ringelmann Effect and Brooks's Law: the Optimal Size of Teams and Agent Squads under Coordination Overhead},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}