Thinking like an architect, in brief
This track is the third rung of Clawdemy’s agentic ladder. Building with Claude taught you to make the parts work: API calls, tools, the agent loop. AI Agent Teams let you read a real multi-agent system from the outside. This track teaches the layer in between, the design judgment that turns parts you understand into a system you would trust in production.
Lesson 1 deliberately contains no code. It establishes the mental frame the other seven lessons build on.
The capability: after this lesson, you can look at a proposed agentic system and ask the architect’s questions: where does judgment live, what is guaranteed versus merely requested, and does this need an agent at all?
What the lesson covers. Every agentic system reduces to four parts: a model, tools, a loop, and a context window. Architecture is how those parts are arranged, and above all, where decisions live. The lesson then walks the field’s most useful dividing line, drawn by Anthropic’s engineering team: workflows follow predefined code paths, agents direct their own. Simpler is better until the path through the work genuinely cannot be known in advance.
The heart of the lesson is three trade-offs that recur in every project: what the model decides versus what code guarantees; one agent versus a team; what stays in the context window versus what gets fetched. The rest of the track takes these up one at a time, from team-scale Claude Code configuration through orchestration and reliability to a build-and-defend capstone.
Why this order. Concept before code is how every Clawdemy track opens, and here the concepts are the product: the tools will churn, the judgment transfers.