Further Reading
A curated reading list for engineers who want to go deeper on software dark factories — primary sources, technical breakdowns, critical perspectives, and practical guides for moving up the Five Levels.
The canonical resources for understanding and implementing software dark factories. Organized by purpose.
Start Here: The Primary Sources
The factory itself: factory.strongdm.ai The mothership. Principles, Techniques, Products. Read in order.
The manifesto: The StrongDM Software Factory: Building Software with AI Justin McCarthy’s original blog post, February 6, 2026. The technical architecture in McCarthy’s own words.
The framework: The Five Levels: From Spicy Autocomplete to the Software Factory Dan Shapiro’s post that established the vocabulary. Read before the manifesto.
The Best Secondary Analysis
Simon Willison’s breakdown: simonwillison.net/2026/Feb/7/software-factory/ The most technically rigorous secondary analysis. Willison on the unit test gaming problem, DTU feasibility, and the cost model. Required reading.
Shapiro on StrongDM: You Don’t Write the Code. You Don’t Read the Code Either. Dan Shapiro’s follow-up, where McCarthy admits he initially doubted the Digital Twin Universe was possible.
Critical Perspectives
Legal accountability: Built by Agents, Tested by Agents, Trusted by Whom? Stanford Law’s CodeX center. The liability gap when no human reviews the code. Essential counterpoint.
The slop question: Dark Factory AI Review Questions whether AI-generated production code constitutes innovation or a degradation of quality.
Technical Deep Dives
Attractor specification: github.com/strongdm/attractor Three markdown files. Read the spec itself to understand the DOT-graph architecture.
CXDB: github.com/strongdm/cxdb The demo product. A real codebase produced by the factory.
Leash: github.com/strongdm/leash Agent policy enforcement. Open source, Apache-2.0.
Specification Writing
Addy Osmani’s spec guide: How to Write a Good Spec for AI Agents The canonical practitioner guide. Goal-oriented language, three-tier permission model (Always/Ask/Never), LLM-as-judge, conformance suites, CLAUDE.md best practices. Read this before writing your first NLSpec.
Simon Willison: Designing Agentic Loops: simonwillison.net/2026/Jan/22/designing-agentic-loops/ How to structure the human-agent feedback loop at each level. Practical patterns for escalation, confirmation gates, and recovery.
AGENTS.md standard: agents.md The cross-vendor spec file. Start here before you write your first NLSpec.
For Engineers Leveling Up
darkfactory.dev: darkfactory.dev Community tracking resource for the dark factory ecosystem — tools, implementations, case studies, and practitioner discussion.
HackerNoon dark factory pattern: The Dark Factory Pattern Practical framework for moving from AI-assisted to fully autonomous.
SWE-bench leaderboard: swebench.com Current model performance on autonomous coding tasks. Updated continuously.
Latent Space podcast: latent.space Weekly deep-dives on AI engineering. Covers agent frameworks, infrastructure, and factory-pattern implementations.
Benchmarks and Research
METR study (2025): The randomized control trial showing experienced developers took 19% longer with AI tools while believing they were 24% faster. The baseline for measuring your actual progress.
Harvard employment study (2025): 9-10% drop in junior developer employment within six quarters of AI adoption. The talent reckoning is documented, not theoretical.
The Open-Source Stack
For engineers building their own factory:
| Layer | Tool | Link |
|---|---|---|
| Agent spec | Attractor | github.com/strongdm/attractor |
| Spec-first IDE | Kiro | kiro.dev |
| CLI coding | Aider | aider.chat |
| IDE agent | Cline | cline.bot |
| Autonomous | OpenHands | openhands.dev |
| API mocking (DTU approx.) | WireMock | wiremock.io |
| PR review | CodeRabbit | coderabbit.ai |
| Policy enforcement | Leash | github.com/strongdm/leash |
| Context store | CXDB | github.com/strongdm/cxdb |