Notes on product management, AI, deep learning, robotics, and consumer technology — and what it takes to bring frontier technology from concept to scale.
The tasks humanoids will perform well first are not the most annoying ones. They're the ones with high repeatability, structured environments, and a low cost of mistakes — which means humanoids will likely not cook your dinner anytime soon. A working parent's mental model for humanoid adoption.
Read more →
The core PM job is the same: turn model capability into product value. But the product judgment required is different. Here is a side-by-side breakdown of what PMs own across problem definition, inputs, evals, failure modes, and launch readiness — in traditional ML systems versus foundation-model systems.
Read more →
A model can look great in a notebook, perform well on benchmarks, and still fail in production. Not because the research was bad — but because the model was only one part of the system. A framework for how Product, Research, and ML Engineering each own a distinct and essential lens.
Read more →
Foundation models don't replace traditional AI wholesale — they change where intelligence lives inside the system. Drawing on lessons from Alexa, Providence, and Codec Avatars, a framework for knowing which layers to hand to a foundation model and which must stay specialized, deterministic, or governed.
Read more →
After overhearing one too many dinner table conversations about models and architectures, my kids got curious. Here's how I explained neural networks — using cats, dogs, and the brain's tiny helpers.
Read more →
AI is not just a new feature set on top of software — it is a different computational paradigm. Traditional software encodes rules; AI encodes behavior indirectly through data, optimization, and context.
Read more →