01 — FLAGSHIP PIPELINE
A fully automated TDD pipeline for WordPress and Elementor. Write a spec, get verified implementation. Eight sequential phases — each one a mandatory gate requiring 100/100.
Sequential pipeline phases
Gate threshold — no exceptions
LLM providers, automatic fallback
Automated — runs while you sleep
WHAT IT IS
DarkFactory implements the full neurosymbolic TDD pipeline — probabilistic AI generates, deterministic gates verify. A spec goes in one end. Verified, tested, production-ready code comes out the other.
The system works overnight. You submit a spec before you sleep. In the morning, your implementation has passed eight sequential gates — or it hasn’t shipped at all.
"Not an AI assistant that writes code. A system that proves the code works against a verified specification."
HOW IT WORKS
The pipeline begins with a machine-readable spec. An AI agent reviews it for completeness, scope, and unambiguous failure modes. No spec, no pipeline entry.
A second, independent agent examines the spec. No shared context with the writer. Disagreement on scope or completeness triggers a loop — not a pass.
Tests are written against the spec before any implementation exists. Every test maps back to a spec requirement. Tests must fail at this stage.
An adversarial agent reviews the tests for coverage gaps, incorrect assertions, and spec misalignment. The same independent-agent pattern as spec review.
Implementation is written to make the failing tests pass — not to produce plausible-looking code. The tests are the only specification of correctness.
Code review against spec and tests. Does the implementation make the tests pass, or does it work around them? Security and maintainability checks also run here.
Final adversarial review of the complete, tested implementation. This reviewer has seen nothing before — clean eyes, full context, full spec.
System-level adversarial review: does the integrated implementation break any existing contracts? Regression coverage, API compatibility, and cross-feature interaction checks.
9-PROVIDER FALLBACK
DarkFactory routes across 9 LLM providers with automatic fallback. If a model is rate-limited or unavailable, the pipeline continues on the next provider. No manual intervention, no lost runs.
Supported providers include Anthropic Claude, Google Gemini, OpenAI, and others. Provider configuration is declarative — switch profiles without touching pipeline logic.
sarolta
DarkFactory is available as a managed service and as a self-hosted pipeline. Start with a conversation — we’ll assess your spec, your workflow, and how to get verified output into your hands.
