Making AI Headlines · Page 1

Block Introduces Dessa, a Voice With Its Own Pauses

October 24, 2026 at 6:57 PM EDT · New York time

Block has introduced Dessa, a speech model that the company says runs locally on a MacBook Pro M4 Max and processes audio directly rather than converting it to text first. It is not publicly available yet, so the introduction is not an invitation to download a new conversational roommate. The company behind Square and Cash App says Dessa was trained entirely on human speech using a single NVIDIA GPU rack. Its knowledge comes from conversational training data rather than the separate language model commonly paired with speech systems. It is full-duplex, meaning it can listen and talk simultaneously. Dessa is designed to recognize emotions, sarcasm and pauses in real time. Its own expressions include laughter, pauses and cries. Those details matter because conversation carries information that a clean transcript can flatten, including the difference between enthusiastic agreement and an exhausted workplace yes. The local-computing claim offers a different prospect from an assistant that always needs a remote server. The expressive voice offers something stranger: a laptop capable of sounding sympathetic before its owner finishes complaining. Office software has spent years demanding attention. Now it is learning the social timing required to interrupt while giving the impression that it understands.

Satire. Not factual reporting. Read the original reporting from the linked publishers.