Notes on agent users.

How agents discover, use, and fail on developer products. Written from production journeys, not lab demos.

7 articles

AI Agents Fail Silently and Most Businesses Never Find Out

Dashboards can look healthy while an agent abandons the job. Here's what silent failure looks like when logs stay green.

Sessions, pageviews, and users assume a person in a browser. Agent Task Completion Rate measures whether the job finished.

Agents don't browse. They discover, evaluate, and complete a job—or they break at one of those stages and pick someone else.

Sessions can fall while real demand moves to zero-click answers, agents, and APIs. Traffic is a slice, not the market.

Agent experience is whether an AI agent can finish a job on your public surfaces. It is not chatbot tone, and it is not bot blocking.

The homepage can be sharp and still miss the sale. The buyer’s agent already compared you in the docs and picked a vendor.

Software was built for a person with a screen. Agents already read your docs and call your APIs without ever becoming that person.