Vibe Coding: Foundations (1 of 2) — GDG Build with AI
Vibe Coding: Foundations introduced a practical framework for directing AI coding tools, with explanations intended for both developers and people new to programming. The session established shared vocabulary around repositories, pull requests, tests, agents, and sandboxes, then traced a five-stage progression from manual coding to parallel agents working in the cloud.
The central idea was to define intent before generating code: what the application should do, why it matters, what is outside its scope, and how to recognize a successful result. A calendar application served as a recurring example, connecting written requirements to observable actions such as creating an event and receiving an invitation.
The Plan, Work, Review, Compound loop gave that approach a repeatable structure. The talk covered small, reviewable changes; tests that capture expected behavior; checking the application itself; and saving lessons in project instructions so future work benefits from earlier mistakes.
Discussion also examined confident but incorrect output, lost context, review backlogs, missing verification, and exposure of secrets. Production features involving accounts, payments, personal information, or destructive changes were highlighted as areas requiring careful review. Optional preparation for the next workshop was to choose a project, create an instructions file, and try the loop before returning for a hands-on design and build session.