The context

How to build an Agent policy

Agents need a policy. Don’t reinvent governance. Apply what you already do with people: review, limit, escalate, grow trust. Reality begins not when the model speaks, but when it starts doing.

💡

Credit: prompted by Jean Latiere’s work on agents in FinOps (link).

When the defaults no longer fit

FinOps often feels simple on paper, then stubbornly complex in practice. When the defaults stop fitting, the friction isn’t failure. It’s experience. This field note is about what happens when judgement, context, and communication matter more than clean models.

AWS feels different. Here’s how that changes my approach.

Something has changed in how AWS absorbs mistakes. Benevolence hasn’t disappeared, but it’s no longer where many practitioners expect it. This essay examines what happens when discretion moves into systems, contracts become the floor, and old assumptions stop holding.

How to build an Agent policy

Read next

When the defaults no longer fit

AWS feels different. Here’s how that changes my approach.

Why Shift-Left often just shifts the load