Founder and CEO of HumanLayer, pioneering production-grade AI agent systems with built-in human oversight. Creator of the 12-Factor Agents framework for building reliable autonomous systems in complex codebases.
Building the Future of Agent Infrastructure
Dex founded HumanLayer (YC F24) to solve a critical challenge in AI agent deployment: ensuring human judgment and safety control remain in the loop when deploying autonomous systems in production. His work focuses on context engineering—the architectural discipline of managing how AI agents process and reason about large, complex codebases.
The Context Engineering Pioneer
Before HumanLayer, Dex spent years helping software teams deliver Kubernetes applications at Replicated, where he developed deep expertise in distributed systems and developer tooling. He also built AI agents that managed SQL warehouses, gaining firsthand experience with the challenges of deploying autonomous systems at scale. This background informed his insight that production AI agents fail not from lack of model capability, but from poor architectural decisions around context management.
The 12-Factor Agents Framework
Dex created the 12-Factor Agents framework, which has gained significant traction in the AI engineering community (tracking toward 10k GitHub stars). The framework addresses the core problem: AI agents operate on limited context windows while production codebases exceed those windows. His solution emphasizes:
- Quality over quantity of context - Strategic information architecture beats raw token budgets
- Multi-agent architecture - Small specialized agents with focused context outperform single large agents
- Context compaction - Compress information to markdown summaries while preserving intent
- The 40% rule - Keep context utilization under 40% to avoid diminishing returns
- Intentional work cycles - Plan, execute, compress, refresh, and repeat
HumanLayer
HumanLayer is a Y Combinator-backed platform (F24) that enables organizations to deploy AI agents with built-in human oversight and control. The platform addresses the critical challenge of maintaining human judgment in the loop while scaling agent deployment across 100K+ developers. HumanLayer provides structured approval workflows, context management for complex systems, and multi-agent orchestration patterns that integrate with existing developer tools and CI/CD processes.
Conference Appearance
Event: AI Engineering Code Summit 2025 Date: November 21, 2025 Time: 9:31 AM - 9:50 AM Session: “No Vibes Allowed: Solving Hard Problems in Complex Codebases”
Dex presented on context engineering as the fundamental constraint in AI agent systems. His talk introduced the concept of “Context Engineering (CtxE)” as a core discipline for building production-ready agents, explaining why quality context beats quantity, how the “40% dumb zone” affects model performance, and why multi-agent architectures outperform single large agents. He emphasized that production agent deployment requires treating AI development with the same rigor as critical infrastructure: explicit planning, context compaction, and human oversight at critical decision points.