Neither edge of a context ACK is a spoken-word boundary

Neither edge of a GPT-Live context acknowledgement is a reliable spoken-word boundary. The model can begin a correct reply before the estimated context interval starts, so replacing an end cutoff with a start cutoff still rejects valid speech. Keep the correlated ACK as an authorization condition, but do not discard a current reply's bounded caption prefix solely because it precedes that interval. Associate evidence with the current directive and interruption epoch, then validate its content separately. This is a correction to the earlier start-cutoff recommendation, not proof of audible delivery or of a particular buffering implementation.

Applies when: GPT-Live output monitoring combines context acknowledgements and approximate transcript intervals. Keywords: OpenAI GPT-Live context acknowledgement start end transcript prefix timing