note

Keeping understanding in the loop

An open question in my work with agents: how do we carry the reasons forward, along with the results?

Working with agents raises a question I keep returning to: how do I retain understanding as more of the implementation is handled for me?

When I return to a project, I need more than the result. I need the decisions, what led to them, what was tested and what is still uncertain.

It shows up in different forms across my work: an owner inspecting an Octops proposal, a handoff after a technical change, or project context that survives a conversation. I’m still working through what helps.

I want agents to expand what I can do while leaving me able to understand and direct the work. Keeping a record is a start. Being able to examine it, ask a question and challenge a decision matters too.