关于「创始人是一台单服务台:排队论与切换成本」,《冲刺转巡航与三小时拍板制》给出的定量结论是什么?

结论:第三章把创始人当作单服务台,利用率从八成升到九成五时等待放大 4.75 倍;单一三小时窗口把最长整块创作时间从 3.34 小时提到 5.68 小时。

数据口径:本书检验这一说法,结论是它在条件变化之后并不成立:当自动化覆盖率、例外率、现金跑道与事故恢复时间越过准入线,继续冲刺的边际产出不为正,把本人最高杠杆投入收敛到一个有限窗口(示意三小时)不会让长期产出更差。 书中参数为示意;全部计算由书中程序生成;精力与睡眠只作自我追踪趋势;本书不是医疗建议,也不是个人化经营或投资建议。

English version

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: 3.34, 5.68

A. 712 requests over 60 days
policymean waitP90 waitstarts/daycreative hlongest block
always0.120.1310.09.283.34
two5.3615.343.510.884.36
one8.8720.112.311.175.68
查看完整验算程序
# Chapter 3. p32: answer-at-once versus fixed approval windows.
# 60 synthetic days, 14 waking hours each (08:00 to 22:00). Decision
# requests arrive as a Poisson stream, 12 a day; service is exponential
# with mean 12 minutes. Each time the desk starts work after being idle
# the creator also loses a residue of 15 minutes of creative focus
# (illustrative, after the task-switching literature). Policies:
#   always   serve on arrival, first come first served
#   two      two windows, 11:00-12:30 and 17:00-18:30
#   one      one window, 14:00-17:00
# Urgent requests (5 %) are served at once under every policy.
# Unfinished work carries to the next window. Reported: mean and P90
# wait in hours, desk starts a day, creative hours left, and the mean
# longest uninterrupted creative block.
import math
import random
import statistics

DAYS = 60
RATE = 12
ES = 0.2
RESIDUE = 0.25
URGENT = 0.05
POLICIES = {"always": [(0.0, 14.0)],
            "two": [(3.0, 4.5), (9.0, 10.5)],
            "one": [(6.0, 9.0)]}


def arrivals(rng):
    out = []
    for d in range(DAYS):
        t = 0.0
        while True:
            t += rng.expovariate(RATE / 14.0)
            if t >= 14.0:
                break
            out.append((d * 24.0 + t, rng.expovariate(1.0 / ES),
                        rng.random() < URGENT))
    return out


def in_window(t, wins):
    d = int(t // 24.0)
    for a, b in wins:
        lo, hi = d * 24.0 + a, d * 24.0 + b
        if lo <= t < hi:
            return True
    return False


def next_open(t, wins):
    d = int(t // 24.0)
    best = None
    for dd in (d, d + 1):
        for a, b in wins:
            lo = dd * 24.0 + a
            if lo > t and (best is None or lo < best):
                best = lo
    return best


def simulate(reqs, wins):
    events = sorted(reqs)
    t, i, last_end = 0.0, 0, -1.0
    queue, waits, busy = [], [], []
    starts = [0] * DAYS
    while i < len(events) or queue:
        while i < len(events) and events[i][0] <= t:
            queue.append(events[i])
            i += 1
        pick = None
        for q in queue:
            if q[2]:
                pick = q
                break
        if pick is None and in_window(t, wins):
            for q in queue:
                pick = q
                break
        if pick is None:
            nxt = []
            if i < len(events):
                nxt.append(events[i][0])
            if queue:
                nxt.append(next_open(t, wins))
            t = min(nxt)
            continue
        queue.remove(pick)
        if t > last_end + 1e-9:
            starts[min(int(t // 24.0), DAYS - 1)] += 1
        waits.append(t - pick[0])
        busy.append((t, t + pick[1]))
        t = t + pick[1]
        last_end = t
    return waits, busy, starts


def creative(busy, starts):
    hours, longest = [], []
    for d in range(DAYS):
        lo, hi = d * 24.0, d * 24.0 + 14.0
        segs = sorted((max(a, lo), min(b, hi)) for a, b in busy
                      if b > lo and a < hi)
        used = math.fsum(b - a for a, b in segs)
        hours.append(14.0 - used - RESIDUE * starts[d])
        gaps, cur = [], lo
        for a, b in segs:
            gaps.append(a - cur - RESIDUE)
            cur = max(cur, b)
        gaps.append(hi - cur)
        longest.append(max(gaps))
    return statistics.mean(hours), statistics.mean(longest)


if __name__ == "__main__":
    rng = random.Random(40432)
    reqs = arrivals(rng)
    print("A. {} requests over {} days".format(len(reqs), DAYS))
    print("   policy  mean wait  P90 wait  starts/day  creative h"
          "  longest block")
    for name in ("always", "two", "one"):
        w, b, st = simulate(reqs, POLICIES[name])
        w.sort()
        p90 = w[int(0.9 * len(w)) - 1]
        ch, lb = creative(b, st)
        print("   {:<6}  {:>8.2f}  {:>8.2f}  {:>10.1f}  {:>10.2f}"
              "  {:>12.2f}".format(name, statistics.mean(w), p90,
                                   statistics.mean(st), ch, lb))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《冲刺期向巡航态平滑切换、每日三小时高杠杆拍板制与创作者永续生命力母典》第 3 章「创始人是一台单服务台:排队论与切换成本」(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-404 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026404,
  author    = {Fan, Yafu},
  title     = {Transitioning from Sprint to Sovereign Cruise: The Three-Hour Daily Decision Cap and Sustainable Creator Vitality},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论