Courage Garikai › Publications
Working Paper 2026-02
Simulated Completeness
Claim–mechanism gaps, mechanism audits and the rebuild decision in an AI-era trust product
The paper defines simulated completeness and distinguishes it from prototypes, minimum viable products, technical debt and fraud. It codes the 65 audit findings into a taxonomy of eight claim–mechanism gap types plus hygiene: 47 (72%) were gaps, including every critical and high finding. It also proposes six mechanism probes, a rebuild decision rule, an evidence contract for agent-built work and seven testable propositions. Evidence is the public commit history, the itemised audit, the production database and the service's own usage meter. It is a single pre-revenue case, so its propositions are offered for testing, not as established findings.
Abstract
Generative AI has made the visible surface of a software product (interfaces, dashboards, documentation, screenshots, test badges, release notes) far cheaper to produce than the mechanisms that make the surface true. This paper names the resulting condition simulated completeness: an artefact whose surface signals operational completeness that its mechanisms do not support. It reports a design-science case study of AttestOS, a bot-verification service whose first version, built in one day in October 2026, looked finished. Its owner console showed traffic, revenue and connected payment webhooks. Its portal showed customer usage, its scanner reported vulnerabilities and its outreach engine drafted security notices.
A read-only audit the next day logged 65 findings, 18 of them critical. Seven ways around the live security engine were confirmed against production. Every console figure was hard-coded. Nobody could obtain a key or pay. A plugin loaded its script from an unregistered domain, and the drafted security notices derived their findings from a hash of the recipient's domain name.
Coding the findings shows that 47 of the 65 (72%) were claim–mechanism gaps: a surface asserting something the mechanism did not do. Every critical and every high-severity finding fell into this class; none of the ordinary engineering defects did. The gaps sort into eight types, and each type is exposed by a different probe, not by the venture's own tests. The venture was rebuilt rather than repaired. The rebuild bound claims to mechanisms through single sources of truth and tests, and replaced synthetic outreach with outreach built only from observed facts. Even so, six of the eleven issues found in the following day's checks were small claim–mechanism gaps in documentation and status labels.
The paper contributes: - a definition of simulated completeness that separates it from prototypes, technical debt and fraud; - a taxonomy of claim–mechanism gaps; - a six-probe mechanism audit; - a rebuild-or-repair test; - seven propositions about how simulated completeness arises, where it does most harm and how it can be governed.
The case is a single, pre-revenue venture analysed by an interested party. The propositions are offered for testing.
Contents
- Introduction
- Literature review
- Method and case
- Findings
- Frameworks and propositions
- Discussion
- Conclusion
- References