Part V — Intermediate Track: From Playground to Real Projects

What this part is for

Part V is where you stop building “playground demos” and start building real projects: repos with structure, repeatable environments, stable model-call boundaries, logging, error handling, and versioned prompts.

The main goal is to make your AI work exportable, reproducible, and maintainable—so it survives beyond the first successful chat run.

The shift: from “prompt output” to “product artifact”

In the playground, the model’s text is the artifact. In real software, the artifact is:

• code in files,
• tests that define behavior,
• configs that are safe and reproducible,
• logs that make failures diagnosable,
• prompt files and schemas that are versioned like code.

This part shows you how to run your vibe loop so each iteration produces a stable artifact, not just a good-looking response.

What you’ll be able to do after Part V

• Export from AI Studio into a clean repo with a repeatable local dev loop.
• Build a minimal request/response wrapper architecture that isolates model calls from app logic.
• Handle timeouts, rate limits, invalid outputs, and safety blocks without “random failures.”
• Move from free-text outputs to schemas and validation (no more parsing vibes).
• Design tool/function calling workflows that are safe and bounded.

Prerequisites (recommended)

You’ll get the most out of Part V if you’re comfortable with:

• the vibe loop and “spec first” habits (Part I),
• prompting constraints and diff-only changes (Part III),
• model selection, prompt structure, context management, and safety basics (Part IV).

Ship points (how you keep momentum)

Part V uses ship points to keep you moving without breaking things. Typical ship points in this part:

• SP1: the repo exists and runs locally with a single command.
• SP2: you have a stable model-call wrapper boundary.
• SP3: failure modes are handled (timeouts, rate limits, invalid output, blocks).
• SP4: prompts and schemas are versioned and logged.

The rule is simple: don’t proceed to the next chapter until your current ship point is real (tests or repeatable commands), not just “it worked once.”

Part V map (Sections 13–16)

• 13. Exporting From AI Studio to a Codebase
• 14. Building Your First Useful AI App (Project 1)
• 15. Structured Output: Stop Parsing Vibes
• 16. Tool Calling and Function Calling (Where Apps Become Real)

Where to go next

• Start with 13. Exporting from AI Studio
• Reference: context management