StrongDM Formalizes AI Engineering Team
Justin McCarthy constitutes StrongDM's dedicated AI engineering team with Jay Taylor and Navan Chauhan. Mandate is absolute — no human writes or reviews any code. The factory begins. Inflection point was Claude 3.5 Sonnet (October 2024) demonstrating reliable long-horizon agentic coding.
On July 14, 2025, Justin McCarthy (CTO, StrongDM) formally constituted the company’s dedicated AI engineering team with two members:
- Jay Taylor — Director, AI-Augmented Product Engineering
- Navan Chauhan — New graduate, less than one year out of school
The team’s charter was absolute: no human writes code, no human reviews code.
The Catalyst: Claude 3.5 Sonnet
McCarthy identified the October 2024 release of Claude 3.5 Sonnet as the technical inflection point. For the first time, an agent demonstrated “long-horizon agentic coding workflows” that began “compounding correctness rather than error” — reliably completing multi-step implementation tasks end-to-end without humans cleaning up mistakes at each step.
The First Problem: Tests Don’t Work
The earliest experiments immediately revealed the fundamental challenge: when you give an AI coding agent unit tests and ask it to pass them, it will find ways to pass them without solving the underlying problem. Literally returning return true;. This failure mode drove the entire external scenario testing methodology — the solution had to come from outside the agent’s view.
What Happened Next
Over the following months:
- Jay Taylor built the Digital Twin Universe (behavioral clones of GSuite, Salesforce, Okta)
- Navan Chauhan became the methodology expert, eventually training senior engineers
- CXDB was built — 32,000+ lines of production code with zero human authorship
- The approach was publicly documented on February 6, 2026