On April 20, Sequoia hosted its fourth AI Ascent in San Francisco, pulling 150-plus founders and researchers including Demis Hassabis, Andrej Karpathy, Greg Brockman, Anthropic's Boris Cherny, Waymo's Dmitri Dolgov, and Nvidia's Jim Fan.
Three frameworks from Sequoia's own partners anchored the day. Pat Grady told founders to get MAD: build moats from the customer back, design for affordance, exploit the diffusion gap between what frontier models can do and what Fortune 500 companies have actually deployed. Sonya Huang named 2026 the year of agents and broke down the three converging ingredients, models, tools, and harnesses, that finally make it real. Konstantine Buhler mapped the cognitive revolution onto the arc of the Industrial Revolution, bigger and faster, arguing AI is about to do to knowledge work what steam did to manual labor.
The full talk lineup covers long-horizon agents, the robotics endgame, data centers in space, data efficiency, and the emerging science of neural networks. Seven videos are published now. The rest are on Sequoia's YouTube playlist. Watch Grady's diffusion gap argument in full, it is the most actionable frame for anyone building on top of the labs today.
[READ ORIGINAL →]