Go from a 3-file Python repo to a deployed NBA AI prediction bot in 4 hours. FastAPI + Anthropic Claude + live odds data — built live, shipped live.
AI tools can generate code faster than ever. But speed without understanding is dangerous. Developers who skip architecture, ignore blast radius, and blindly accept AI output are building systems they can't maintain.
This workshop teaches the missing layer: the engineering discipline that makes AI-generated code production-ready. You're not learning tricks. You're building architectural literacy for the AI era.
A closed-loop system for making safe, verified changes to production codebases with AI.
Read nba_bot.py, trace its 3 classes, and find the entry point — before touching a single line. Understand the repo before you change it.
Identify that fetch_today_games() returns hardcoded data — that's target #1. Classify the change, estimate blast radius, find the real fix.
Load the repo architecture into every Claude prompt. Never ask cold — context about nba_bot.py's structure goes in before any code request.
Read every diff before accepting it — the AI will do more than you asked. See exactly what changed in nba_bot.py before committing a line.
Apply only what you asked for — nothing extra. Wire up The Odds API, build the FastAPI endpoint, or add Claude predictions one scoped step at a time.
Run the Production Preflight — 16 items — before every commit. No NBA prediction ships without passing the gate.
Every step in the workshop reinforces this cycle. No step gets skipped.
Scope the problem and set boundaries
Map the codebase in VS Code
Feed context to Claude Code
Inspect every diff in Fork
Clean commit and push
Watch a preview of the live session — see the tools, the workflow, and how it all comes together.
Nathan has spent years building production systems and integrating AI into real engineering workflows. This workshop distills that experience into a repeatable framework you can use on day one.
Learn more about Nathan →Workshop-aligned articles on AI-augmented engineering, production discipline, and the future of development.
AI writes code faster than ever. But the skill it's quietly eroding is the one that matters most.
A structured approach to prompting that keeps AI grounded in your codebase reality.
The mental model every developer needs before making changes to production systems.
75 seats. One Saturday. The engineering discipline that separates professionals from prompt-and-pray developers.