Part 5

The Evidence

A method induced from practice should explain the engineering history that produced it. This Part follows DocAble — a production document-accessibility system built largely by directing coding agents — from a five-minute feasibility experiment to a deployed service. Observing one system from the inside and from the beginning lets us reconstruct what a finished architecture cannot: which pressures appeared, what response followed, what survived, and what had to be revised again.

The clean method arrived last. As the system grew, missing representation became expensive in some places; missing authority became dangerous in others; repeated operational surprises exposed properties nobody had modeled at all. Some obligations were encoded before failure. Some models arose from clean design choices and simply held. Others were forged in incidents and hardened through recurrence. Part V preserves those differences rather than forcing every event through the finished theory.

Depth and breadth

MAGE emerged from one deeply observed production build. That case supplies chronology, mechanism, and within-case recurrence.

Independent industrial accounts supply a different kind of evidence: variation across systems built by other organizations under different constraints.

The first shows how the method emerged. The second asks how far its engineering grammar travels.

New here: Originating case · Support ratio · Delegation staircase · Within-case evidence · Industrial reconstructions · Comparative evidence

Part V uses two views of the evidence; Figure 5.0-1 sets them side by side.

Two views of the evidence A two-column diagram converging to one point. The evidence splits into an originating case, DocAble, offering high process depth, longitudinal history, incidents and responses, and within-case recurrence, which together answer how the phenomenon emerged; and industrial reconstructions, independent systems, offering greater variation, selected mechanisms, different constraints, and alternative realizations, which together answer how far it travels. The two columns converge on analytic generalization, the constructs and mechanisms worth explaining, which motivates the theoretical account developed in Part VI, A Theory of MAGE. Neither view supplies causal or population-level estimates. Two views of the evidence Depth versus breadth — together they motivate the theory in Part VI. THE EVIDENCE ORIGINATING CASE DocAble INDUSTRIAL RECONSTRUCTIONS independent systems HOW DID IT EMERGE? HOW FAR DOES IT TRAVEL? ANALYTIC GENERALIZATION PART VI A THEORY OF MAGE high process depth longitudinal history incidents + responses within-case recurrence greater variation selected mechanisms different constraints alternative realizations constructs + mechanisms worth explaining Neither view supplies causal or population-level estimates.
Figure 5.0-1. Two views of the evidence. The originating case supplies longitudinal depth: sequence, mechanism, and within-case recurrence. Independent industrial reconstructions supply variation across systems and organizations but less process visibility. Together they motivate the theoretical account developed in Part VI.

DocAble supplies sequence, mechanism, and within-case recurrence, but not causal or population-level estimates. The six industrial reconstructions supply variation and alternative realizations, but less process history. Together they motivate rather than establish the theoretical account developed in Part VI.

Parts II–IV presented the compressed method. Here the direction reverses. Chapters 5.1–5.4 return to the originating case from which much of that terminology was induced, so the wrong turns matter. A finished architecture shows what exists; a longitudinal case can show why it exists and what it replaced. Chapter 5.5 then asks whether independently built systems expose comparable structures. Depth supplies mechanism; breadth supplies variation. Part VI asks what general account can explain both.

© James C. Davis, 2026–present