Tech Industry
The gap between companies that understand AI-native operations and those that don't is widening. Here's how I think about where the industry is headed—and what it means for the people building inside it.
Product & Engineering Philosophy
How I think about building products when the building tools change faster than the products themselves.
I believe in three principles that most product organizations claim to follow but systematically violate:
Ship the learning, not the feature. In an AI-native company, the most valuable output of any product cycle isn't the feature itself—it's what you learned about how your users and your AI systems interact. If you're not instrumenting for that learning from day one, you're building blind.
Design for the capability curve, not the current capability. Every AI system you deploy today will be significantly more capable in 12 months. If your product architecture assumes today's limitations are permanent, you'll spend more time refactoring than building. I design systems with explicit upgrade paths for AI capability improvements.
Measure outcomes, not outputs. The number of features shipped, tickets closed, or models deployed tells you nothing about whether you're winning. I build measurement frameworks around customer outcomes and operational efficiency—the things that actually compound.
The Inflection Point
Essay
We are at a specific moment in the history of technology companies, and most people in the industry are misreading it.
The common narrative goes like this: AI is a new tool. We integrate it into our existing products and processes. We become more efficient. The org chart stays roughly the same, just with fewer people in some boxes and AI in others.
This narrative is wrong in a way that will be obvious within three years but is costing companies enormously right now.
What's actually happening is a phase transition in how organizations can be structured. For the first time, the coordination costs that justified large, hierarchical organizations are dropping faster than the benefits of scale. This means the optimal company size for any given market is shrinking—rapidly.
The companies that understand this are not "adopting AI." They are rebuilding themselves around a fundamentally different set of assumptions about how many people you need, what those people should be doing, and how work flows between humans and machines.
I've seen this from the inside. The companies doing it well share three characteristics: they treat AI as infrastructure rather than a feature, they hire operators who can redesign systems rather than just run them, and they measure success by outcomes per employee rather than revenue per product.
The inflection point isn't coming. It's here. The question is whether you're building for the world on this side of it—or still optimizing for the one that's disappearing.