Block's voice AI learned to talk without ever reading a word

Block's tiny research lab unveiled Dessa on October 23, a speech-to-speech model it says was trained from scratch on nothing but recorded human conversation — no text corpus, no pretrained model underneath. Because it learned language by listening, it listens and talks at the same time: it cuts in, backchannels, laughs mid-sentence, and hears sarcasm, sighs, and voice cracks that text-based pipelines throw away.

Block says the model runs locally on a MacBook Pro M4 Max or better, faster than real time, and was trained on a single NVIDIA GB200 NVL72 rack. For now it's a research preview: a public waitlist is open, but early access goes first to a small group of AI safety researchers.