Tolan builds a voice companion around memory

In a case study published by OpenAI, Tolan describes a voice companion built with GPT-5.1. The product combines low-latency responses, real-time context reconstruction, and memory that shapes later conversations. The intended experience is continuous: users should not need to reintroduce themselves every time they return.

Context

For a voice product, delay and memory are part of the interface. A familiar conversational style, however, is not evidence of human understanding. The practical questions concern which information persists and whether a person can inspect, correct, or remove it. Those controls matter alongside the quality of the voice.