Pattern track of the AI series. Adds the primitives that turn an LLM from a text generator into an agent: defining tool schemas, single and multi-step tool calling, chain-of-thought, few-shot CoT, chained prompts, classify→branch routing, deterministic output validation, self-critique, moderation, eval criteria, agent loops, multi-tool agents, eval suites, iteration via evals, retry-on-bad-output, structured typed outputs, multi-step planning chains, agent + Composio tool integration. Each lesson demonstrates ONE primitive on the smallest possible input — minimal generic problems, never products. 30 lessons across 4 weeks. Requires AI Foundations completed (calling LLMs, structured output, multi-turn). Max tier (Sandbox + OpenRouter, with Composio available for the final-week tool integration).
Week 1
Week 2
Week 3
Week 4