Here is the map I use when working with AI agents.#
The Agentic Engineering Periodic System.
For months, I’ve been collecting and studying principles, patterns, and anti-patterns for working with AI agents.
I’ve compared sources and tried approaches in my own work.
That collection grew into the Periodic System.
It organizes practices, their relationships, and the checks that go with them.
Trying these approaches helped me see how the practices fit together.
For me, it became a map.
It helps me with questions that arise every time I work with agents:
What needs to be clear before a task starts?
What context does the agent need?
What can it decide on its own?
What evidence do I need before accepting the result?
Each practice comes with a concrete check you can apply to your own work.
For example, “Evidence before done”: What evidence supports the claim that a task is finished?
The system is freely accessible.
You’ll find the link in the first comment.
Take a current task and choose a relevant practice.
Which of these four questions do you find hardest to answer?
#SoftwareCraft #AI
