batuta falsify
The falsify command runs the Popperian Falsification Checklist - a 108-item quality assurance protocol based on Toyota Production System (TPS) principles and the scientific method.
Usage
# Run full checklist on current directory
batuta falsify .
# Run on a specific project
batuta falsify /path/to/project
# Output JSON format
batuta falsify . --json
# Critical checks only (fast mode)
batuta falsify . --critical-only
Overview
The checklist implements Sir Karl Popper’s falsification principle: every claim must have explicit rejection criteria. Each of the 108 items is a falsifiable claim about the project’s quality.
Sections
The checklist is organized into 10 sections:
| Section | Items | Focus |
|---|---|---|
| 1. Sovereign Data Governance | 15 | Data residency, privacy, consent |
| 2. ML Technical Debt Prevention | 10 | CACE, entanglement, dead code |
| 3. Hypothesis-Driven Development | 13 | Reproducibility, baselines, statistics |
| 4. Numerical Reproducibility | 15 | IEEE754, cross-platform determinism |
| 5. Performance & Waste Elimination | 15 | PCIe rule, SIMD, latency SLAs |
| 6. Safety & Formal Verification | 10 | Memory safety, fuzzing, Miri |
| 7. Jidoka Automated Gates | 10 | CI/CD circuit breakers |
| 8. Model Cards & Auditability | 10 | Documentation, provenance |
| 9. Cross-Platform & API | 5 | Linux/macOS/Windows, WASM |
| 10. Architectural Invariants | 5 | YAML config, pure Rust testing |
TPS Grades
Results are graded using Toyota Production System terminology:
| Grade | Score | Meaning |
|---|---|---|
| Toyota Standard | 95-100% | Production ready |
| Kaizen Required | 85-94% | Acceptable with improvements |
| Andon Warning | 70-84% | Issues require attention |
| Stop the Line | <70% | Critical issues block release |
Severity Levels
Each check has a severity level:
- Critical: Blocks release if failed
- Major: Requires remediation plan
- Minor: Should be documented
- Info: Informational only
Example Output
╔═══════════════════════════════════════════════════════════════════╗
║ POPPERIAN FALSIFICATION CHECKLIST - Sovereign AI Protocol ║
╚═══════════════════════════════════════════════════════════════════╝
Project: .
Evaluated: 2025-12-11T12:00:00+00:00
Grade: ◐ Kaizen Required
Score: 88.9%
Items: 84/108 passed, 0 failed
─── Jidoka Automated Gates ───
✓ JA-01 Pre-Commit Hook Enforcement [MAJOR]
✓ JA-02 Automated Sovereignty Linting [MAJOR]
✓ JA-03 Data Drift Circuit Breaker [MAJOR]
...
✅ All critical checks passed - Release allowed
Integration with CI
Add to your CI pipeline:
- name: Quality Gate
run: |
batuta falsify . --json > falsification-report.json
# Fail if critical checks fail
batuta falsify . --critical-only || exit 1
TPS Principles Applied
The checklist embodies Toyota Way principles:
- Jidoka: Automated gates stop on quality issues
- Genchi Genbutsu: Evidence-based verification
- Kaizen: Continuous improvement through feedback
- Muda: Waste detection and elimination
- Poka-Yoke: Error-proofing through constraints
Related Commands
batuta stack quality- Stack-wide quality metricsbatuta analyze- Project analysis