OpenAI has quietly rolled out a series of significant updates to its ChatGPT mobile application, introducing a novel gesture for fine-tuning AI responses and enhanced navigation features for extended conversations. These improvements aim to give users more control over the AI's output and streamline the user experience.
New Gesture for Prompt Intelligence Control
A "secret gesture" has been added to the ChatGPT app, allowing users to precisely adjust the effort level or "intelligence" the AI dedicates to a prompt before generating a response. To access this feature, users simply need to press and hold the send arrow within the chat interface. This action reveals an intelligence-level selector, offering options such as Instant, Thinking, and Extended modes.
The selection dictates the depth and detail of the AI's reasoning, empowering users to request anything from quick, concise answers to comprehensive, elaborate explanations. It's important to note that the availability of all intelligence-level options may vary based on the user's OpenAI subscription plan.
Improved Navigation with Table of Contents
For users engaging in lengthy discussions, OpenAI has implemented a practical table of contents feature. When a conversation spans five or more responses, ChatGPT automatically generates distinct sections based on the topics discussed. A menu icon, positioned on the right side of longer conversations, becomes visible. Hovering over this icon reveals the table of contents, enabling users to quickly jump between different parts of the discussion, significantly improving navigability.
Currently, this table of contents feature is available on the ChatGPT web application, with potential rollout to the mobile app in future updates.
Under-the-Hood Model Enhancements
Beyond the new interface features, OpenAI has also made adjustments to the underlying GPT-5.5 Instant model. These changes are intended to "improve response style and quality," according to the company, suggesting a continuous effort to refine the AI's performance and conversational abilities. Users may notice subtle differences in response quality, speed, reasoning, or overall behavior following these model updates.