Hard Gates across the Client Lifecycle: the Decay of Soft-Prompt Compliance over Turns and State Machines for Interception before Publication: chapter answers
People who work through AI agents often believe that if reminders are clear enough and layered enough, violations will approach zero. This book tests that belief with the mathematics of reliability. Soft reminders work in proportion; when layers are independent the residual violation rate is the product of their miss rates, and with illustrative parameters four layers still leave 0.0048, not zero. Only a deterministic code gate can reach zero, and only inside the pattern class it defines. After checking the sources the book corrects its founding blueprint: Reason's Human Error of 1990 writes…
DOI pending · monograph page goes live after publication · 5DL-B-256 · 中文版
- Chapter 1 · The Product Account: Why Stacked Soft Layers Never Leave a Zero Residual
Soft reminders work in proportion; stacked, they leave a product, and with illustrative parameters four layers leave 0.004800, not zero. A hard gate sorts violations into an in-class part, which escapes only when the ga… - Chapter 2 · Holes That Line Up: the Floor That Common-Cause Failure Puts under Stacked Soft Layers
Chapter 2 treats common-cause failure: with a common-cause share of only 0.01, it supplies 0.933 of the residual after six soft layers. - Chapter 3 · Decay over Turns: an Exponential Model of Soft-Prompt Compliance, Re-Sending Periods and Reading the Log
Chapter 3 gives an exponential decay model with a floor, with a half-life of 6.93 turns, under which a forty-turn session lets through 4.71 times the static estimate. - Chapter 4 · Where to Intercept: before the Write, after It or before Publication, and Whether a Broken Gate Passes or Refuses
Chapter 4 compares interception before the write, after it and before publication; the combined design E costs only 0.60 at a hook failure rate of 0.010. - Chapter 5 · Exhausting the State Machine: Proving a Dirty Publication Unreachable and Finding the Shortest Counterexamples of Three Faulty Designs
Chapter 5 proves by exhausting a state machine that a dirty publication is unreachable under the right design; with a green light not bound to content, 0.1964 of event sequences of length ten go wrong. - Chapter 6 · The Price of Hardness: Pattern Width, Wrong Blocks and Gates That Get Switched Off
Chapter 6 prices wrong blocks: a wide rule set of coverage 0.97 survives only 15.52 weeks a year and lets through 354.64 violations. - Chapter 7 · Classical Sources East and West: Law and the Inked Line in the Han Feizi, and Reason's Classification of Error
Chapter 7 examines the chapter "You Du" of the Han Feizi and Reason's classification of error; behind the gate, 0.627 of the residual is knowledge-based mistakes. - Chapter 8 · Closing the Loop: the Weekly Interception Report, the Residual Calculator and a Pre-Registered Test
Chapter 8 gives a weekly interception report (in an invented example the block rate falls from 42.39 to 12.71), the Gate Residual Violation Rate Calculator and a pre-registered test (power 0.916). Marked boxes separate…
Illustrative model calculation, not a live quote