I'm David Vogel. I help founders and engineering teams ship AI that actually holds up in production. Most of what I do looks unglamorous: deleting frameworks, clarifying context, tightening loops, and writing things down so the system stays legible.
My work is built on Interpreted Context Methodology — the folder-structure approach to agentic workflows pioneered by Jake Van Clief. ICM replaces opaque orchestration libraries with a primitive every engineer already understands: the directory. It’s the most reliable way I’ve found to build agent systems that survive a quarter, let alone a year.
I administer a premium Skool community of practitioners — reserved for serious operators confirmed to be working at this level. The community is the closest thing to a residency in production-grade AI.
Before this, I shipped AI features in environments where downtime was measured in revenue and observability mattered more than novelty. That bias still drives everything I build.