ChatGPT for iOS: Interactive Forms in Codex Tasks
OpenAI's ChatGPT for iOS app (version 1.2026.195) added support for interactive forms directly inside Codex tasks, letting developers provide structured input to a running task without leaving the transcript view. The same release also improved task-list sorting by recent activity, added selected-text previews in the composer, and let users resume blocked or usage-limited goals runs. Together these changes make managing Codex work from a phone more fluid, particularly when a task needs a quick decision or input rather than a full desktop session.
Key Takeaways
- Interactive forms now let Codex tasks collect structured input (selections, confirmations, short fields) directly inside the ChatGPT iOS transcript, instead of relying on free-text replies.
- This closes a desktop-to-mobile parity gap: structured decision points in a task were previously harder to act on quickly from a phone.
- The feature builds on inline visualizations, added just a week earlier (2026-07-13), continuing OpenAI's pattern of enriching what a Codex task transcript can render on mobile.
- Task lists now sort by recent activity, making it faster to find the task that needs attention among several running in parallel.
- Goals runs can now resume after being blocked or paused by usage limits, reducing the need to restart a long-running autonomous objective from scratch.
- Third-party coverage of this specific point release is minimal; it shipped as one line item in OpenAI's changelog rather than a standalone announcement, consistent with how OpenAI typically ships incremental ChatGPT iOS updates for Codex.
Sources & Mentions
1 external resource covering this update
Interactive Forms in Codex Tasks
ChatGPT for iOS now renders interactive forms inline inside Codex task transcripts. When a Codex task needs structured input from the user, such as selecting an option, confirming a choice, or filling in a short set of fields, the app can present that input as a native form rather than requiring the user to type a free-text reply. This closes a gap between the desktop and mobile Codex experience: previously, tasks that paused for structured input were harder to act on quickly from a phone, since the user had to interpret the request from plain text and reply manually.
This builds on OpenAI's broader push to make Codex fully usable from mobile. Earlier iOS releases in July 2026 added the ability to create, search, fork, and manage Codex tasks directly from a ChatGPT conversation, and to view inline visualizations such as charts and diagrams inside task transcripts. Interactive forms extend that same "act on a task without opening a laptop" philosophy to structured decision points, not just passive viewing.
Task Management and Composer Improvements
The release also refined how the app surfaces ongoing work. Task lists now sort by recent activity, so the most relevant in-progress tasks surface first, and the app shows a clear indicator when a host machine is unavailable while starting a new task. The composer gained selected-text previews, letting users see and re-reference a highlighted snippet after it has been sent, along with smoother transitions when starting a new task.
Codex's goals feature, which lets developers set a higher-level objective for Codex to work toward autonomously, gained the ability to resume runs that were previously blocked or paused due to usage limits, rather than requiring the user to restart the goal from scratch.
Bug Fixes
The update also addressed several stability issues, including an app crash that could occur when duplicate entries appeared in the task list while starting a task, incorrect task-list styling and task actions on iOS 18, and spacing issues in the composer and attachment menu.