Summary
In this episode of the a16z Show, partners discuss the fast-moving world of consumer AI agents and personal assistants. The conversation starts with a recent agent meetup of about 100 builders, where ten assistant startups pitched and founders were notably open with one another about a landscape no one yet knows how to map.
A central theme is trust. Drawing on an ongoing benchmark of more than 147 assistants, the guests note that roughly 22% of user messages about fears center on trust. They explore how assistants differ along axes from serious and productive to playful and even adversarial, and why paternalistic products tend to fail.
The discussion also examines onboarding and context gathering. Examples include prompts that ask users to export data from ChatGPT, voice calls that interview the user, and photo-based context from products such as Dazzle, from Marissa Mayer’s new company, which emphasizes forgetting unneeded photos. The hosts compare these approaches with screenshot-based personal data and debate how much access people will give.
The episode offers a useful view into how builders and investors think about personalization, privacy, memory, and the emerging ecosystem of everyday AI agents.
📺 Source: a16z · Published November 04, 2026
🏷️ Format: Podcast
