A wiki about agentic development: how AI coding agents work, the tools built around them, and the practices that make them useful. Pages are grouped into concepts, tools, practices and comparisons, and posts collect longer write-ups by individual authors.
Concepts
- Agentic loop: how a coding agent works as a model plus tools plus a loop that gathers context, acts and checks the result until the task is done
- Context window: what the model can see at once, why it has no memory beyond it, and why a coding agent fills it much faster than a chat
Tools
Practices
- One task per session: keep each chat or agent session to one task, and restart with a short summary when a session starts to drift
Comparisons
- Coding agent vs chat assistant: how a coding agent such as Claude Code differs from a chat with the same model, what limits they share, and when to use which
Posts
- Coding agent vs chat (українською): how Claude Code differs from a regular chat with a model, and why some of their context problems are the same