The Five Shifts
To internalize the architecture mindset, you need to make five specific shifts in how you think. These will recur throughout the course. By the capstone, they should be instinctive.
Shift 1: From prompting to designing. A prompt is an instruction. A design is a system. When you encounter a complex task, stop asking “what prompt would produce the best output?” Start asking “what architecture would produce the best output consistently, repeatedly, and with improving quality over time?”
Shift 2: From breadth to depth. A single generalist agent covers everything shallowly. A specialized agent covers one thing deeply. When you’re tempted to make one agent do everything, ask yourself what it would miss that a specialist would catch. The answer reveals where you need another agent.
Shift 3: From output to flow. The individual agent’s output matters less than the information flow between agents. A brilliant analysis that the next agent in the pipeline ignores is worth nothing. When you evaluate your system, look at the transitions, not just the outputs.
Shift 4: From static to evolving. A prompt sits in a text file, unchanged. An agent evolves through operational feedback. It gets better at what it does because production teaches it what the paper design couldn’t anticipate. When you design an agent, don’t aim for perfection. Aim for a v1.0 that’s honest about its gaps and structured to evolve.
Shift 5: From single to system. Stop thinking about “my agent.” Start thinking about “my system.” An agent is a component. The system is what produces value. A brilliant agent in a bad system produces less than a decent agent in a great system.
You have just read the five shifts. Reading them and internalizing them are different operations. The following self-assessment asks you to rate where you currently stand on each shift. There are no correct answers. The purpose is to create a baseline you can revisit after the capstone to measure how your thinking has changed.
Your self-assessment is a snapshot, not a score. Some of these shifts will feel intuitive after Module 3. Others will not click until Module 8, when you build your first multi-agent system and experience the difference between output thinking and flow thinking. Hold onto your baseline. The capstone will ask you to revisit it.