Making AI Deterministic for Developers and their Agents, with Patrick Vuong of Moderne

In this episode of Code Story, host Noah interviews Patrick Vuong, Director of Product at Modern, a company pioneering deterministic tooling for AI agents in software development. Patrick shares his journey from Microsoft to Modern, highlighting how his experience in developer tooling and agentic workflows shaped his passion for building reliable, scalable AI systems. Modern’s core mission is to transform how AI agents operate by providing deterministic, context-rich tooling that overcomes the limitations of probabilistic, hallucination-prone models. The company’s approach centers on a 'lossless semantic tree'—a compiler-accurate representation of code—that enables agents to reason with precision and efficiency. Products like TriGrep, Prethink, and ChangeLog empower agents to search faster, access pre-computed context, and avoid redundant work, all while reducing token usage and improving reliability. Modern positions itself as a 'deterministic harness' that bridges the gap between agent-driven innovation and traditional SDLC governance, enabling autonomous 'dark factories' where agents operate at scale with human oversight. The episode underscores a critical shift: AI's future isn't just about smarter agents, but about building the right infrastructure to make them predictable, auditable, and cost-effective at enterprise scale.