Block Unveils Dessa, a Full-Duplex Speech Model

Block has published an announcement for Dessa, a speech model it describes as "the most authentic full-duplex speech model ever built."

The term full-duplex means the model can listen and speak at the same time, rather than taking strict turns the way familiar voice assistants do. The approach is meant to mirror natural human conversation, where people interrupt one another, slip in short replies and react without waiting for a sentence to finish. That sense of "naturalness" is exactly what the headline leans on.

Beyond that, the available material offers little: Block's announcement page could not be loaded, so the model's technical specifications, availability and intended uses remain out of view.

The interest in the topic shows indirectly through the audience response: on Hacker News the post gathered 186 points and 80 comments. That is a notable, if not record-setting, level of attention — the sort of reception typical for releases that spark lively discussion without yet becoming a major event.

Without the text of the announcement itself, it is too early to judge how well the claim to be the "most authentic" model holds up. Bold phrasing of this kind is common in conversational AI, and only hands-on testing can settle it.

Source: Hacker News