Sara Hooker Keynote at PyTorchCon 2026 on Adaptive Intelligence
The keynote reframes frontier AI progress around adaptation efficiency rather than pretraining scale, pointing practitioners toward test-time scaling and adaptive compute techniques they can apply now. Developers and researchers optimizing agents should watch the AutoScientist work on harness optimization, memory and dataset creation. The 2026 prediction that individuals will build and own frontier AI signals a shift in who controls state-of-the-art models.
At PyTorchCon 2026, Sara Hooker (@adaption_ai) examined the limits of scaling larger models and the shift toward adaptive intelligence. Her talk covered post-training, test-time scaling, adaptive compute and hardware co-design as alternatives to simply scaling pretraining. She highlighted AutoScientist examples spanning agent harness optimization, agent memory and dataset creation. Her keynote argued that making AI easier to adapt and customize could let more people build and own frontier AI.
At #PyTorchCon 2026, Sara Hooker @sarahookr (@adaption_ai) examines the limits of scaling larger models and the shift toward adaptive intelligence.
Sara discusses post-training, test-time scaling, adaptive compute, and hardware co-design as alternatives to simply scaling pretraining. She highlights AutoScientist examples spanning agent harness optimization, agent memory, and dataset creation.
Her keynote explores how making AI easier to adapt and customize could enable more people to build and own frontier AI.
2026 prediction? "You will build and own frontier AI"