Agentic Programming with Codex

What if you could take an app from design mockup to a fully tested, working application on desktop and web — without writing the code yourself?
In this session we'll do exactly that.

Using OpenAI's Codex as our coding agent, we'll walk the complete loop live:

Figma design → components → database-backed feature → correction → cleanup → tests → running app

Our example is a Kotlin Multiplatform contact list app, and Codex does the heavy lifting at every stage:

• Pulling design context straight from Figma via MCP to generate pixel-accurate UI components
• Building the full feature — ViewModel, repository, SQLDelight database, validation, and screens — from a single well-scoped task
• Taking a focused correction when it over-engineers (it will), and fixing only what's broken
• Running automated cleanup to strip dead code and enforce architecture rules
• Generating unit tests to full coverage on models and ViewModels

By the end you'll have seen the same app running on desktop and in the browser, built almost entirely by the agent