
The Startup Moat Moved: What Every AI-Native Founder Has to Unlearn
Three months as an EIR at Andrew Ng's AI Fund — on finding the pain, the moats that are left, and how you actually build now.
The “why” is curiosity — never challenge. It's how I get to root cause.
Founder, Exponential OS — a multi-agent harness with a constitution, agentic memory, model routing and composable skills. 25+ years in engineering. 6 years at Google across a ~$40B portfolio, $500M+ ROI. Most recently Engineer in Residence at AI Fund, Andrew Ng's venture studio.
As an Engineering Leader, I have found that the most complex technical challenges—from managing $40B portfolios at Google to compressing supply chain recall from months to seconds—are solved by identifying the core “Why.”
To be clear: this is a strictly curious “Why”. It is never about challenging authority or arrogance. It is a relentless drive to drill down to the fundamental root cause and fix the broken system.
The method is to build the harness before the product: governance that enforces engineering invariants as gates, memory that carries context across sessions, and evals fixed before any code is written. A change ships only if it beats baseline. That is how I ship outcomes rather than features.
Question the baseline before you scale.
Design for reliability at every failure mode.
Outcome-driven engineering, not just feature delivery.
Architecting reliability-first AI systems and evaluation-driven agent platforms. ex-Google engineering leader.
View Case StudiesFounder, ExponentialOS. Patent filed (exponential OS architecture). 10x Hackathon Champion. Compressing 6-month cycles to 6 seconds.
View Case StudiesBerkeley Haas EMBA 2026. UC Berkeley Faculty — 1,500+ Fortune 500 executives taught AI systems. Leading 50+ engineers at scale. $40B Google portfolio.
View Case Studies
Three months as an EIR at Andrew Ng's AI Fund — on finding the pain, the moats that are left, and how you actually build now.

Ranking on Google and being citable by AI are not the same problem — and I asked four engines the same question to prove it.

What the ClawCamp talk was really about, and the patent-night story I didn't tell on stage.
Notes from Health+Tech 2026 conference session on agents, attention, and what survives the transition
Architectural Integrity • Global Operations • AI Governance