AI Agents is a practical library for learning delegated software without the hype. An agent is a program that pursues a goal on your behalf: it breaks work into steps, uses tools, checks its own results, and asks for help when it should. That is genuinely new, and it changes how work gets assigned, reviewed, and trusted. These guidebooks are for the people living through that shift — readers who want to understand what agents can and cannot do before handing one a calendar, a codebase, or a customer.
The library moves from concepts to operations. The early guides define the territory and explain the machinery. The middle of the shelf is about working alongside agents day to day: writing tasks they can succeed at, setting permissions, reviewing output, and diagnosing failures. The deeper entries read like an operations manual — evaluations, sandboxes, incident response, prompt injection, rollback — for teams running agents in production rather than trying one at a demo.
Where to start
Read What AI Agents Are: The Moment Software Started Taking Initiative first for a clear definition and the loop that makes agents different from chatbots and scripted automation. Then move to How to Delegate to AI Agents: A Playbook for Better Tasks to learn how to hand over work, and AI Agent Permissions: The Ladder From Read to Act before you grant anything the power to send, spend, or delete. When something goes wrong — and something will — When AI Agents Fail: How to Debug the Delegation shows how to find the real cause. The full reading path lives on the guidebook shelf . For hands-on play, solve embodied-agent puzzles in Hot Swap or use the AI Agents Learn with Fizz track to drill the core guidebook ideas.
The editorial approach is evidence-first: claims are tied to how these systems actually behave, limits get as much attention as capabilities, and safety is treated as a design habit rather than an afterthought.
