Limitless: An AI Podcast We Interviewed The Team Behind OpenAI's #1 Feature
14 snips
Oct 6, 2025 Christina Kaplan and Samir Ahmed from OpenAI lead personalization efforts for ChatGPT, focusing on innovative features like Pulse. They dive into how ChatGPT has evolved from a basic assistant to one that prepares your day while you sleep. Discussion includes the transformative April memory update, short- versus long-term memory, and user behavior changes, as well as privacy concerns. They clarify that ChatGPT serves as an assistant rather than a digital clone, and explore exciting future capabilities that could seamlessly integrate various aspects of users' lives.
AI Snips
Chapters
Transcript
Episode notes
Human Memory As A Design Lens
- OpenAI frames memory using human cognitive metaphors but acknowledges AI memory is far less faithful than human memory.
- Memory aims to meet users where they are and normalize human-style interaction with models.
Give Context Once, Reap Future Value
- Share persistent context proactively so ChatGPT can use it later instead of only asking isolated questions.
- Provide facts upfront (e.g., dietary preferences) to get tailored recommendations over time.
Sensitive Data Enables Practical Value
- Personal data uploaded to ChatGPT enables high-value, sensitive uses like medical prep, but requires strong privacy and security.
- OpenAI views ChatGPT as potentially the most sensitive personal account and treats data protections seriously.