AI AGENTS
MASTERCLASS

15+ frameworks, one mental model

Every framework is a different harness around the same loop: reason, act, observe, repeat. Learn the model once and read them all.

The framework landscape — mapped by autonomy and orchestration style.

Six capabilities of a real agent

An assistant answers. An agent reasons, acts, and self-corrects toward a goal. These are the six muscles that make the difference.

ReAct vs ReWOO — animated

The two dominant loops. ReAct interleaves thought and action; ReWOO plans everything first, then executes. Watch both run.

ReAct — reason, act, observe, repeat.

ReWOO — plan upfront, then execute the plan.

The whole masterclass, in motion

Eighteen diagram GIFs and seven terminal walkthroughs. No static slides — every idea animates.

mega-agents-everything — the full picture in one reel.

Start with zero API keys

The minimal ReAct agent runs out of the box. Then graduate to LangGraph, CrewAI, the OpenAI SDK, and typed Pydantic AI agents.

# A tiny ReAct loop — reason, act, observe, repeat
def run(goal):
    state = {"goal": goal, "scratch": []}
    while not done(state):
        thought = reason(state)        # LLM picks next action
        result  = act(thought.tool)    # call a tool
        state["scratch"].append(observe(result))
    return finalize(state)

if __name__ == "__main__":
    print(run("What's 17% of 2,480?"))

Running in three commands