Andrew Feldman made the remarks onstage at TechCrunch Disrupt in San Francisco on October 13. Asked what must happen for AI to become a household appliance, he said it is not far away. He expects assistants to work partly on the device and partly in the cloud, like a phone, to know their owners well, to operate inside a security perimeter the user sets, and to hold payment cards so they can buy things on the user's behalf. Feldman said he already uses a set of agents to tidy his schedule and brief him on upcoming callers. He prefers infrastructure and left the applications to other founders. In a separate session he argued that the industry's focus on AI risks is drowning out its benefits. For builders, the hybrid pattern points to split inference, with small local models for private context and cloud models for heavy work. Delegated payments and permission boundaries are the product problems to solve first. Feldman sells inference chips, so he gains if agent usage grows.