
Enterprise AI’s 200-Millisecond Problem
Your model will take 20,000 tokens. Choosing which 20,000 is a structure problem, and the unit worth modeling is the decision trace.
A human-first community newsletter exploring the future of context engineering and AI. One story, one lesson, and one shared insight at a time. Curated with 💙 for the humans of data & AI.
Live Last read · last published · next check

Your model will take 20,000 tokens. Choosing which 20,000 is a structure problem, and the unit worth modeling is the decision trace.

Storing your context in a graph can be the right call. Hooking that graph up to the agent almost never is, and the two decisions get made as though they were one.

314 obstacles across ten reports, and a vacancy sitting between two org charts

You cannot inventory agents fast enough to govern them. You can govern the parts they are made of.

Adoption numbers tell you people are using AI. They don't tell you if the company is getting more capable.

Why context portability is the threat most AI programs cannot see yet

A walkthrough of my talk at the AI Engineer World’s Fair: two eras of building agents inside my own company, five stack migrations, and the infrastructure question I left the room with.

The metrics that look green on the dashboard are the ones hiding where your system actually fails. Aggregate metrics show the average and conceal what matters most.

Why the accountability gap in your AI deployments was decided at the go-live celebration

What it takes to build the learning loop Satya described, and why most organizations leak it before it compounds.

Your AI’s context has a floor. Most teams have never measured theirs. Here’s the one number I use.

Nobody owns the Hard Half. Technical context had a stage. Organizational context didn't have a booth.