Breaking

Murati Says Next-Generation AI Will Listen Continuously and Reply in 20-Millisecond Slices

November 8, 2026 at 11:15 EST

Foundation Models

Mira Murati has said that today's AI models are effectively "deaf and blind" while they reason, and that the next generation will instead process input continuously and respond in chunks of about 20 milliseconds.

Mira Murati · AI models

Today's AI is "deaf and blind" while it thinks. Next-gen models may never stop listening.

Murati describes a future where models keep taking in audio and other signals while they reason, and respond in chunks of about 20 milliseconds instead of waiting for a full turn.

20 ms

Size of each output chunk the future model would emit

Chunks of 20 ms that fit inside one second (1,000 ms ÷ 20)

Product names, release dates or benchmarks disclosed so far

One second, drawn to scale

Each square is one 20 ms chunk. The orange square is a single output chunk.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.