Insights
Thoughts and perspectives on AI, technology, and professional development spanning 20+ years of experience
Published on July 9, 2026
Summary
Drawing on a centuries-old family wax seal, this article explores what history's original approval gate teaches us about governing AI agents. I examine four failure modes of ungoverned multi agent work — source drift, role collapse, hidden context dependence, and candidate creep — and the harness that contains them: one canonical source, separated roles across separate models, a quarantine for new ideas, and a single human seal.
My Perspective
For most of history, anyone could write, but only one accountable hand could make words official. Working with teams of LLM agents, I learned we are relearning that lesson the hard way. The ability to produce text and the authority to make it true are different powers, and everything depends on keeping them separate. And because every gate, register, and review trail has to persist somewhere, governance is ultimately a storage workload. The seal made the decree, but the archive made the kingdom governable.
This article is hosted on LinkedIn. You may need to sign in to LinkedIn to view the full content.
Published on May 2, 2026
Summary
Drawing on Greg Graffin's book Population Wars, this article explores how his evolutionary-biology argument for coexistence over conquest reframes AI industry competition. I look at NVIDIA's "five-layer AI cake" as a real-world example of resource partitioning: no single company can own the whole stack, and the winners are the ones building ecosystems within their layer rather than trying to dominate every layer.
My Perspective
Graffin's central claim is that species and civilizations that endure do so through coexistence and resource partitioning, not by wiping out rivals. Watching the AI infrastructure industry organize around interdependent layers — energy, chips, infrastructure, models, applications — convinced me the same lesson applies. NVIDIA's partner-led approach to building an ecosystem, rather than trying to own every layer, is coexistence as strategy, not conquest.
This article is hosted on LinkedIn. You may need to sign in to LinkedIn to view the full content.
Published on October 7, 2025
Summary
Drawing parallels between marathon running and AI infrastructure implementation, this article explores how the endurance, preparation, and strategic pacing required in marathons mirror the challenges of deploying robust AI data infrastructure. I examine how DDN's approach to AI storage reflects the same principles that help marathon runners cross finish lines successfully.
My Perspective
Having observed countless organizations sprint at the start of their AI journey only to hit walls later, I've learned that AI infrastructure is a marathon, not a sprint. Just as marathon runners must pace themselves, manage their energy, and prepare meticulously, successful AI implementations require sustained commitment, proper planning, and the right infrastructure foundation. DDN's technology embodies this marathon mindset - built for the long haul, not just quick wins.
This article is hosted on LinkedIn. You may need to sign in to LinkedIn to view the full content.
AI & Technology
7 articles
Enterprise & Cloud
4 articles
Leadership & Innovation
6 articles