You've learned the seven principles of general agent problem solving:
Knowing the principles is one thing. Applying them together in real workflows is another. This lesson shows how the principles combine in practice, with concrete examples you can adapt to your own work.
Here's the mindset shift: You're no longer the one typing code. You're the director managing an agent.
Think of the AI as a junior developer with infinite energy but no institutional knowledge. It will work tirelessly, but it needs:
Your job is no longer typing—it's directing, reviewing, and approving. The seven principles are your management framework.
Real workflows rarely involve a single principle in isolation. They require multiple principles working together:
You get a report: "Users are seeing 'Invalid token' errors when logging in."
Principle 1: Bash is the Key
Principle 7: Observability
Principle 3: Verification as Core Step
Principle 4: Small, Reversible Decomposition
Principle 6: Constraints and Safety
You need to add: "Users can reset their password via email."
Principle 5: Persisting State in Files
Principle 2: Code as Universal Interface
Principle 4: Small, Reversible Decomposition
Principle 3: Verification as Core Step
Principle 1: Bash is the Key
Principle 6: Constraints and Safety
Principle 7: Observability
Refactor src/auth/auth.js (500 lines) into smaller, testable modules.
Principle 5: Persisting State in Files
Principle 4: Small, Reversible Decomposition
Principle 2: Code as Universal Interface
Principle 3: Verification as Core Step
Principle 1: Bash is the Key
Principle 6: Constraints and Safety
Principle 7: Observability
Not all principles are equally important for every task. Use this guide to prioritize:
What happens when step 6 (verification) fails? Don't panic—follow this recovery path:
Key insight: Failure is information. The failed attempt tells you what DOESN'T work, narrowing down what will. Revert cleanly, learn from the failure, try again.
Use this checklist to evaluate how well you're applying the principles:
Count how many principles you're actively applying (1 point per principle with at least 2 checkboxes marked):
Where to start if you're at 0-2: Begin with just two principles—Verification (always test) and Observability (always see what AI did). These two alone prevent most disasters.
The principles are powerful individually. Together, they're transformative:
When you apply all principles together, you move from "using AI" to "collaborating with an intelligent agent." The workflow becomes:
This is the Digital FTEs: Engineering paradigm in action.
All seven principles derive from one meta-principle:
General agents are most effective when they leverage computing fundamentals rather than fighting against them.
File systems, shells, code execution, version control—these aren't limitations to work around. They're the foundations that enable reliable, debuggable, powerful agent workflows.
Claude Code makes this explicit through the terminal interface. Cowork makes it accessible through a GUI. But underneath, they're running on the same principles—the same Claude Agent SDK, the same reasoning engine, the same fundamental approach to problem-solving through computing primitives.
Both interfaces support all seven principles. Choose based on your task characteristics:
The choice isn't "which is better"—it's "which fits this task." Many workflows benefit from using both: Claude Code for implementation, Cowork for documentation and review.
Here's a power move: tell the AI which principle to use.
Instead of vague instructions like "refactor this code," try:
Or for a bug fix:
Why this works: You're giving the AI a framework, not just a task. It knows HOW you want it to work, not just WHAT you want done. This puts you firmly in the director's seat.
What you're learning: How to integrate all seven principles into a cohesive workflow. You're experiencing how the principles work together to make AI collaboration more effective.
What you're learning: How to evaluate and optimize your existing workflows using the seven principles as a framework. You're learning to identify gaps and implement improvements.
What you're learning: How to select and prioritize principles based on task characteristics. You're developing judgment about which principles matter most in different contexts.
These seven principles form the foundation for effective problem-solving with General Agents. In Part 2 and beyond, you'll apply these principles to increasingly sophisticated workflows:
Each workflow will demonstrate how the principles combine in practice, using whichever interface—Claude Code or Cowork—best fits the task. The principles don't change; only their application deepens.