Planning loops, tool use, memory, and multi-agent coordination.
Build agents that plan, use tools, remember, and coordinate — with clear eyes about where they fail and how to keep them under control.
Engineers past RAG and prompt-chaining who need models to do multi-step work, and leaders sizing up what agents can realistically ship today.
The guides are numbered — read in order for the curriculum path, or jump straight to the one you need. Each card is self-contained.
What actually makes an agent an agent, five defining capabilities, and why 2026 is when the pattern lands.
Chatbots, workflows, and agents each solve different classes of problem. How to honestly choose.
Think to act to observe, and ReAct.
How agents act on the world.
Short- and long-term memory design.
Decomposition and re-planning.
Reflection, planning, tool use, collaboration.
A working agent from scratch, no framework.
When many agents beat one, and when they do not.
Reasoning-driven retrieval inside agents.
Standardised ways to connect agents to systems.
Comparing the main build options and trade-offs.
Measuring trajectories, not just outcomes.
The new attack surface and layered defence.