What does Human-AI Symbiosis and Parallel Agents conclude, in numbers, about “How Many Bits an Instruction Must Carry”?
Answer: Chapter 2 counts the bits in an instruction: with 20 attributes that must all be right, 1.0 bit each gives acceptance 0.000484 and 3.0 bits 0.998734; a strong test cuts wrong deliveries from 0.4049 to 0.0134.
Scope of the numbers: Its agents, rates, hours and values are illustrative models; all calculations are generated by its programs; the book is not legal, financial or employment advice and makes no claim about any particular AI product. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses.
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: 0.000484, 0.998734, 1.0, 20, 3.0
| bits per attribute | total bits | P(one ok) | P(all 20 ok) |
| 0.0 | 0 | 0.3829 | 0.000000 |
| 0.5 | 10 | 0.5205 | 0.000002 |
| 1.0 | 20 | 0.6827 | 0.000484 |
| 1.5 | 30 | 0.8427 | 0.032619 |
| 2.0 | 40 | 0.9545 | 0.394018 |
| 3.0 | 60 | 0.9999 | 0.998734 |
| bits to reach MSE 0.01 on one attribute: 3.32 |
Show the full program
# Chapter 2. p03: How many bits must an instruction carry? Rate-distortion
# for a Gaussian source: to bring an attribute with prior spread 1 to
# within mean squared error D takes R(D) = 0.5 log2(1/D) bits. A piece of
# work has k = 20 independent attributes; it is "accepted" only if every
# attribute lands within +-0.5 of the target. With R bits per attribute
# the residual spread is 2^-R, and acceptance is the product over all k.
# A smooth gain per attribute looks like a threshold for the whole.
from math import erf, log2, sqrt
K, TOL = 20, 0.5
def p_within(sd):
return erf(TOL / (sd * sqrt(2)))
if __name__ == "__main__":
print(" bits per attribute total bits P(one ok) P(all 20 ok)")
for r in (0.0, 0.5, 1.0, 1.5, 2.0, 3.0):
sd = 2 ** -r
one = p_within(sd)
print(" {:18.1f} {:10.0f} {:9.4f} {:12.6f}".format(
r, r * K, one, one ** K))
print(" bits to reach MSE 0.01 on one attribute: {:.2f}".format(
0.5 * log2(1 / 0.01)))Single source and citation
Fan, Y. (2026). Human-AI Symbiotic Cognition, Parallel Agent Pipelines and Prompt Information Bounds (Version 1.0.0), Chapter 2: How Many Bits an Instruction Must Carry. 5D Library. DOI pending
DOI pending · monograph page goes live after publication · 5DL-B-103 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026103,
author = {Fan, Yafu},
title = {Human-AI Symbiotic Cognition, Parallel Agent Pipelines and Prompt Information Bounds},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}Other chapter answers in this book
- Chapter 1 · Drift to the Middle, and What Taste Adds
- Chapter 3 · Amdahl, Gunther and Parallel Agents
- Chapter 4 · The Serial Part Is the Commander
- Chapter 5 · Share of Time, Share of Value, and Signing Off
- Chapter 6 · An Instruction as a Noisy Channel
- Chapter 7 · The Huainanzi, Han Xin and Shannon
- Chapter 8 · Practice: A Commander's Week and a New Operator's Training