An AI agent's first feature should be a boundary
Useful AI assistance starts with a bounded job: explicit access, limited authority, durable task state, and evidence that separates an answer from an action.
A PERSONAL PERSPECTIVE ON TECHNOLOGY
Technology is the starting point.
What we do with it is what matters.
I'm Greg. I write about cloud, practical AI, developer productivity, and turning big ideas into meaningful outcomes. A few lessons learned, and plenty still to explore.
NOTES FROM THE JOURNEY
Useful AI assistance starts with a bounded job: explicit access, limited authority, durable task state, and evidence that separates an answer from an action.
A deployment describes what should run. An operating model explains how to change it, verify it, and recover when reality disagrees.
A successful backup job is useful evidence, but recovery depends on readable data, available credentials, a documented sequence, and a working application.
FOLLOW YOUR CURIOSITY
Different threads. A shared curiosity.