Every morning at 8 AM, a developer opens their laptop to find Codex has already organized the day's priorities. Before they can ask "What should I work on today?" the AI has scanned their notes, drafts, and recent activity to produce a ranked task list. This scene, which topped GitHub trending this week, is now a reality for anyone using OpenAI's coding assistant.
OpenAI Academy launched Codex Automations on April 23, 2026
The new feature lets Codex execute tasks on a fixed schedule without manual input. Instead of waiting for a user to type a command, Codex returns to a conversation thread at a preset time, performs the requested work, and leaves the results visible. This marks Codex's transition from a passive response tool to an active agent that initiates work on its own.
A typical use case: "Come back to this conversation every Monday morning and plan my week based on my notes, drafts, and priorities." Codex maintains context within the same thread, so it can pick up ongoing work rather than starting fresh each time. The feature is especially useful for recurring tasks that require continuity across sessions.
To set it up, users define the trigger and the instruction:
bash
Example automation prompt
"Every weekday at 8 AM, review my recent notes and drafts, then create a prioritized task list for today."
Codex runs locally on the user's machine, so the laptop must be on and Codex active for automations to fire. OpenAI Academy recommends starting with a manual conversation to refine the exact behavior and output format before converting the pattern into an automated schedule.
Previously, users had to wake Codex up. Now it returns on its own
The old workflow required opening the chat window and typing "Can you do this for me?" Codex responded once and stopped. The new model inverts that relationship: the user configures a trigger once, and Codex wakes itself at the appointed time, completes the task, and leaves the output ready for review.
This shift from a summoned tool to a self-initiating agent is the core architectural change. For local deployments, automations work best when the laptop remains powered on and Codex is running in the background. The feature does not require cloud connectivity beyond the initial setup.
The immediate developer benefit is time saved on repetitive overhead
Developers often spend 10 minutes each morning reconstructing context: "What did I do yesterday? What needs attention today?" Codex Automations eliminates that ritual. It can generate weekly reports, summarize changes, and check for updates automatically. The user's role shifts from producer to reviewer.
OpenAI Academy defines three criteria for good automations: specific, repeatable, and easy to review. An ideal prompt might be:
"Every Friday, review my recent work and give me a short summary of what I completed this week, what's still open, and what I should watch out for next week."
The recommended workflow is to first iterate with Codex manually until the output matches expectations, then convert the conversation pattern into a scheduled automation. This ensures the AI understands the exact format, tone, and level of detail required before running unattended.
Codex now moves on its own schedule. The question is no longer what the developer asks the machine, but what the machine asks the developer when it returns with results.
