The AI autonomous work window is doubling every 90 days, according to research lab METR, and predicts that three new frontier models will be released in the next 100 days. The first Blackwell-trained base model, named "Spud," completed pre-training on March 24 and is expected to be shipped by June 30. The new model can autonomously identify and exploit software vulnerabilities at a level beyond human intelligence, outperforming all but the most skilled humans. This shift is significant as Google's upcoming I/O event is set to showcase its new Gemini 3 and 4.