Who Is Watching the Agents? Why Finishing Isn't the Same as Success
The agent finished. Did the business recover? This Field Note explores why observability becomes a condition for delegation.
Read field note →Topic archive
Field Notes
The Dashboard Is Losing Its Monopoly
The dashboard is not disappearing. It is losing its monopoly on how we access operational knowledge. When the consumer of that knowledge can also change the environment, we need to observe its decisions as well as the system.
Read field note →Who Is Watching the Agents?
The more authority we give AI agents, the better our evidence needs to be about what they actually accomplished.
Read field note →Maybe One AI Shouldn't Do Everything (Video)
maybe the future is not one incredibly smart AI doing everything. It may be a set of bounded agents with different authority and clear handoffs.
Read field note →How Much Authority Would You Give AI? (Video)
trust should expand in stages. Reading email, drafting email, sending email, publishing content, changing infrastructure, and spending money are different levels of authority.
Read field note →Every AI Agent Needs These 4 Steps (Video)
a serious AI agent needs a loop of Reason, Judge, Act, and Verify. Intelligence is not the same thing as authority.
Read field note →A Successful API Call Can Still Be Wrong (Video)
a successful tool call does not prove that an AI agent made the right decision. Verification is part of the action.
Read field note →Meta Muse Just Made Personal AI Mainstream (Video)
Meta just made personal AI agents a mainstream consumer conversation. The more useful an agent becomes, the more important it is to ask what makes it stop.
Read field note →I Gave My AI Agent More Control. Then I Started Asking Who Could Stop It. (Video)
What makes the agent stop? A Video Note about personal AI, authority, judgment, and verification.
Read field note →