The author, a long-time ChatGPT user, expresses frustration over the lack of capacity warnings and continuity in long-running conversations, which affects their work. They propose specific improvements, including a visible capacity indicator, automatic warnings before reaching limits, user-controlled consolidation of chat history, and the ability to reference previous conversations for better project management. These changes are crucial for users relying on AI for extensive projects.