Baseline: what this repository's rules are compared against

The review is finished when the baseline is named. This file names it and says how a comparison would run.

Named baseline

Checklist: Lampson 1983 and its revision Lampson 2020 (hints, goals STEADY, techniques AID); the alternative checklist is the AWS Well-Architected Framework with six pillars (AWS 2026p).
Check: Newcombe 2015, a precise model of the design checked before code exists.
Taught method: Primer 2026, four steps from use cases to scaling, the most adopted by stars.
Failure benchmark: the failure classes of Yuan 2014 and Alquraan 2018, as the things a design must survive.

This repository's rules are about writing the design down: state load, failure model and consistency assumptions beside the design.
The baseline has no rule about the document; it has rules about the system. The comparison is therefore of what a reader finds, not of a rule list.

Reader tasks: "find what the design assumes about failure" and "find a flaw in a seeded design", on designs with known flaws from Alquraan 2018.
Conditions: the design written with the repository's rules, and the same design written in the primer's four steps.
Design: crossover, so each reader sees both documents on different designs; order balanced.
Measures: flaws found, time, and wrong beliefs about the failure model; effect sizes with intervals.
Control for the reader's experience; report the sample.

No study of design documents for running systems was found, so no effect size exists to power the comparison.