Codex turns demonstrations into reusable AI workflows

Ep 884: Codex Record and Replay: How to Teach an Agent Once Your Most Time-Consuming Workflows (Replay)

Record and Replay points toward agents that learn from observed work rather than waiting for carefully written prompts.

3 key takeaways

1Record and Replay watches a user complete a task, then converts the demonstration into a reusable agent workflow.
2The approach moves AI interaction beyond chat by connecting demonstrations to skill files, video input, and cross-application tasks.
3Reliable automation still depends on security controls, validation, and iterative refinement before workflows can run independently.

Don't miss

The key moment is the reframing of automation as teaching an agent by recording a workflow once, then validating and refining its replay.

The brief

Jordan Wilson introduces OpenAI Codex’s Record and Replay, a feature designed to observe manual work and turn it into an agent-ready workflow.

The episode’s central shift is from prompt-based interaction to demonstration-driven automation: users show an agent what to do instead of describing every step.

Record and Replay extends beyond a single application through skill markdown files, video ingestion, and workflows that move across software tools.

The practical promise comes with a constraint: recorded workflows must be tested, refined, and secured before they can reliably handle time-consuming work.

Weekly email

State of the Pod, every Monday

The week's most-discussed names and stories from podcasts, with the moments behind them. Free, five minutes, unsubscribe in one click.