deepmind

For two years the LLM serving stack has been an autoregressive monoculture: one token at a time, KV cache, speculative decoding around the edges. Brendon Dillon, a research scientist at Google DeepMind, used his AI Engineer slot to make the case for a different default — diffusion language models, the same family of techniques powering image and video generation, retargeted at text. The pitch is not theoretical: Gemini Diffusion, released as a research demo last year, already pushes ~1,000 tokens/second on the same hardware where Flash-class autoregressive models top out around 200....

DeepMind’s roughly six-minute overview video, “Generating novel scientific hypotheses with Co-Scientist,” opens not with a product demo but with a confession from scientists: the firehose of new literature has long since outpaced the humans trying to drink from it. One researcher describes having “hundreds of Chrome tabs and papers open.” Another says the amount of knowledge needed to stay at the frontier of a field now doubles roughly every two months....