How ChatGPT Mobile Redesigned Conversations for Persistence

AI · 7 min read

ChatGPT's shift from sessionized chats to persistent conversation threads changed mental models overnight. The app introduced sticky context, memory toggles, and a redesigned conversation list that surfaces recent topics, with careful affordances to prevent unintended long-term storage of sensitive inputs.

Multimodal inputs (image, audio, and later video) were integrated with progressive disclosure: a single plus button expands into capture options, while inline prompts and annotation tools let users refine AI outputs without leaving the chat. The teardown highlights trade-offs between discoverability and UI complexity.

On the backend, embeddings and retrieval-augmented generation power the 'memory' feature, and we examine cost-performance trade-offs: how context window management, TTL policies, and user-controlled deletion tools maintain compliance and user trust.