Apr 30
Loading video player...
Codex is no longer just a coding assistant in a terminal. In this workshop, Vaibhav and Katia show how it becomes a full software engineering system, combining frontier models, the Codex app and CLI, plugins, automations, and subagents that can explore, review, and execute work in parallel. The session also dives into custom subagents: showing how specialized agents with different models, permissions, and tools can speed up code review, research, debugging, and long-running tasks while keeping control and safety in place. Speaker info: - Vaibhav Srivastav | https://x.com/reach_vb - Katia Gil Guzman | https://www.linkedin.com/in/katiagilguzman/ Timestamps (0:14) Introduction to the workshop and speakers (1:48) What is Codex? (Overview of software engineering agent capabilities) (2:18) Foundation models (GPT-5.3, Spark, GPT-5.4) (4:11) The evolution of models and performance improvements (Websockets, Fast Mode) (7:04) Codex app features, projects, and work trees (8:37) Automations overview (12:28) Plugins: Skills, apps, and MCP servers (15:13) Game and Web development plugins (Playwright and Image Gen) (16:26) Demo: Game studio plugin and generating assets (17:23) Demo: Google Drive plugin for codebase data (18:52) Demo: Setting up automations for Slack and Gmail (27:14) Code Review features and integration with GitHub (32:39) Subagents: Parallelizing tasks and custom personas (36:18) Demo: Using subagents to review persona files (44:52) Creating custom subagents (49:29) Bleeding edge features: Guardian approvals, hooks, and personality settings (56:11) Codex security and Cloud Code plugin (57:10) Q&A session