Full Report
ChatGPT's Codex, which is an AI agent that lets you code and delegate programming tasks, is now testing a new feature that lets you choose the best solution. [...]
Analysis Summary
# Main Topic
Update to ChatGPT's Codex AI agent introducing a new feature to allow users to review and select the best code solution among multiple suggestions for a given programming task.
## Key Points
- Users can now explore multiple code solutions proposed by Codex for a single task and choose the most appropriate one, rather than receiving only one option.
- New keyboard shortcuts have been introduced: `⌘-/` on macOS and `Ctrl+/` on other platforms to open Codex.
- The update includes the addition of a "branch" query parameter, supplementing existing parameters like "environment," "prompt," and "tab=archived."
- Codex is based on `codex-1`, an optimized version of the ChatGPT o3 model, designed for increased coding accuracy.
- Codex functionality includes pulling codebases from GitHub, writing new code, proposing pull requests, and running tasks in isolated sandboxes.
- A setup script duration limit has been increased to 20 minutes for Pro, Team, and Business subscribers.
- Code diff clarity has been improved, allowing users to `option-click` a diff header to expand/collapse all diffs.
## Threat Actors
- No specific threat actors or malicious campaigns mentioned in relation to this feature update. The context focuses on a product enhancement.
## TTPs
- Not applicable (N/A). The content describes development features, not adversary Tactics, Techniques, and Procedures (TTPs).
## Affected Systems
- ChatGPT Codex environment.
- Users with relevant subscriptions (The Codex feature is noted as available with a $20 Plus subscription).
- Systems running macOS or other platforms supporting the new Codex keyboard shortcuts.
## Mitigations
- No specific threat mitigations mentioned as this is a feature release summary.
- **Actionable Information within Context:** Users are advised to utilize the new features, including exploring alternative solutions and using keyboard shortcuts for efficiency.
## Conclusion
This intelligence update details functional enhancements to OpenAI's Codex, moving the AI coding assistant toward a more comparative and efficient workflow by allowing multi-solution exploration. No immediate security threats or actor activity are associated with this specific development. Users leveraging Codex for code generation should familiarize themselves with the new selection mechanism and efficiency improvements.