Will users care whose interests their AI agent serves, or delegate to agents regardless?

Reid Hoffman · Jun 10, 2024

What people call loss of privacy is really a fear of loss of control, and deep data integration will prove worth it as its benefits become concrete.

A biosigns monitor that warned you and a doctor fifteen minutes before a heart attack and dispatched an ambulance would be enormously valuable knowledge.

36:31 20VC: Reid Hoffman on Foundation Models: Who Wins & How Do Incumbents Respond | The Inflection AI Deal: How it Went Down | Why Trump is a Threat to Democracy | The Future of TikTok | Lessons from Sam Altman, Brian Chesky and the OpenAI Board

Harry Stebbings · Sep 5, 2025 · speculative

People would be willing to submit all their data for training and labeling in exchange for about five pounds a month

Real-world data capture and labeling is one of the biggest bottlenecks, and glasses could capture millions of inputs cheaply

Scope: framed as a bet

46:40 20Growth: Meta CMO Alex Schultz on How All Founders Have to Change Their Marketing Playbook in a World of AI | Is AI Plateauing, What it Means if China Wins the AI Race and Why Zuck is a Generational Leader

Harry Stebbings · Jul 6, 2026

Nobody actually cares about privacy other than hardcore technologists

Scope: acknowledges it is blunt and will be unpopular

20:39 20VC: Why Now is the Time for the Application Layer | Why OpenAI & Anthropic Won't Win the App Layer | Why Startups Should be TokenMaxxing | Why VCs Should Reduce Weighting on Price & Ownership in an Age of AI with Mike Mignano, USV

Mike Mignano · Jul 6, 2026

Products and businesses will increasingly push users toward an always-on AI world because context is becoming more valuable to both the labs and the application layer.

Context is increasingly valuable, so companies have incentive to capture more of it; social norms have already shifted, e.g. from discomfort with Alexa's wake word to people casually saying they're 'Granola-ing' a meeting.

11:18 Context value pushes products to always on and norms already shifted

Mike Mignano · Jul 6, 2026

Users will increasingly choose agents and harnesses whose incentives are aligned with them rather than with the labs, and will become self-conscious about who their agent is working for

As you hand an agent your agency, personal information and credit cards, you want it working for you, whereas the incentives of big-lab models are to make the lab's models smarter and better

18:56 Users pick agents whose incentives align with them not the labs

Harry Stebbings · Jul 6, 2026

People will become completely comfortable handing over the keys to agents, and this will be de facto within five years

The same pattern as getting comfortable putting credit cards online, using Apple Pay, and meeting spouses online

20:24 Full comfort with delegation within five years

Mike Mignano · Jul 6, 2026 · hedged

It is possible that users will care about agent incentives this time, because agents act as a second self — buying things and sending personal messages — in a way that previous web technologies never did

Browsing the web let companies learn our interests but we stayed in control; agents will act on our behalf, which may make people think harder about whose incentives the model serves

20:45 Agents as a second self may break the historic indifference