Rebuilding the Stack

Starting a new job means getting into proper volume of work across the modern AI developer toolkit. A clear list of what matters:

• Claude Code and Claude Cowork — for iteration and progress
• Python best practices for modules, packages, app architecture
• NotebookLM as a knowledge management platform
• FastAPI for backend
• Docker containers — which ones to use
• MCP servers — which to leverage
• EC2 + auto-scaling for processing large batches
• CI/CD pipelines — vision and implementation
• Unit, Integration, User Acceptance Tests

Best practice for AI DX workflow: use AI as a partner during all stages — initial ideation, design, documentation, prototype, testing, deployment, troubleshooting, monitoring.

The goal is not to do all of this at once. Life does not end when the job starts. It improves further.

The DX insight here: the stack is not a list of tools. It’s a set of relationships. Which tools do you trust, how do you hand work off between them, and where do you put the human judgment?