Skip to main content

QA Methodology

QA Track in Implementation Phase​

Delivering high-quality microproducts to users requires effort in ideation, implementation, and operation. This guide focuses on the implementation phase, where QA enables fast execution without sacrificing reliability. QA is a dedicated track inside Phase 2, not a post-build afterthought. Its purpose is to create effective feedback loops, prevent regression, and support timely delivery of a product that meets or exceeds initial expectations.

Disclaimer​

Like much of the content in this knowledge base, our QA guidelines and templates do not prescribe a single doctrine or toolchain. Instead, they consolidate lessons from other microproducts into opinionated defaults: guiding principles, reference architecture, and pragmatic best practices. This is a working document, and we welcome alternatives and improvements through contributions.

QA Philosophy​

We draw inspiration from successful engineering organizations and adapt ideas from shift-left testing, continuous testing, and behavior-driven development (BDD) to the realities of modern AI-assisted microproduct development.

Our quality assurance methodology has the following characteristics:

  1. Practical Software Design
  2. Automation
  3. Thoughtful Quality Gates & Feedback Mechanisms

Which are explored further below.

Practical Software Design​

Low coupling, deep vertical slices, and intentional project organization promote effective testing and visibility.

Applying software design principles is critical to making a project testable and allowing it to grow in a way that avoids regressions and remains legible to stakeholders.

Independent components, often organized as vertical slices, make it easier to reason about where business rules and behaviors live in the system. They also help agents build or extend capabilities without affecting unrelated areas.

Building small pieces of functionality end to end also ensures incremental progress and allows testing and feedback at each step. This does not strictly require UI-to-database slices; a helper class with a clear purpose and a test suite that expresses business requirements is equally valuable. Such classes should be treated as first-class components of the system.

Well-designed systems embrace business-domain concepts and terminology in their specifications, tests, and documentation. Dependencies between modules should be apparent to domain experts.

Automation​

Automated tests, CI/CD pipelines, and other AI-assisted workflows facilitate rapid development as the codebase and feature set grow.

Thoughtful Quality Gates & Feedback Mechanisms​

Deterministic checks at specific points in the change lifecycle give useful signals to agents and human reviewers, balancing correctness with development friction.

There are several good points in the change lifecycle to install quality checks, including:

  • Agent loop (run on each message or code change)
  • Human QA check (feature branch)
  • Pre-commit/pre-push hooks
  • Pull request
  • CI/CD pipeline
  • Release

Static analysis tools, tests, and automated code review are invaluable for keeping project quality and entropy in check. The right toolset and enforcement points vary by team and project, but choosing them deliberately is part of the transition from prototype to product.

Agent-Assisted Testing Guidance​

  • Prioritize tests for the highest-value workflow and failure modes first.
  • Use known sample datasets to validate deterministic expectations.
  • Require human review for ambiguous or high-risk agent-generated changes.

Repository Entropy Controls​

  • Enforce consistent architecture and naming conventions.
  • Gate merges with required checks and explicit reviewer accountability.
  • Track defect recurrence and unstable areas to prevent drift.

Risk-Tiered Checklists and Release Gates​

Apply checks by risk tier:

  • Low risk: functional path test and basic regression check.
  • Medium risk: error-path, integration, and accessibility spot checks.
  • High risk: full critical-path validation, rollback readiness, and monitoring checks.

Release Readiness Checklist​

  • Critical user path passes in representative conditions.
  • Known high-impact defects are resolved or explicitly mitigated.
  • Observability is in place for new high-risk code paths.
  • Rollback and incident response owner are identified.

Contributor Expansion Areas​

  • QA automation patterns for agent-first repositories.
  • Risk scoring models for release decisions.
  • Post-release quality review templates.

Propose an improvement