StrongDM built a "Software Factory" where AI agents autonomously write, test, and validate code without human intervention. The approach emerged after Claude 3.5's October 2024 update enabled long-horizon coding that compounds correctness rather than errors. Key innovations include replacing traditional tests with "scenarios" (end-to-end user stories validated by LLMs), measuring success through "satisfaction" scores rather than boolean pass/fail, and creating a "Digital Twin Universe" - behavioral clones of third-party services like Okta and Slack for unlimited validation. The team operates under strict rules: humans must not write or review code, with a target of $1,000+ daily token spend per engineer. This represents a fundamental shift in software economics where previously infeasible approaches (like building full SaaS replicas for testing) become routine.
Table of contents
The StrongDM AI StoryFind Knobs, Turn To ElevenFrom Tests to Scenarios and SatisfactionValidating Scenarios in the Digital Twin UniverseUnconventional EconomicsRead NextSort: