‘Hard Fork’ Live, Part 1: Satya Nadella and Cindy Cohn

What It Covers

Hard Fork Live features Microsoft CEO Satya Nadella and outgoing EFF Executive Director Cindy Cohn in live conversations covering AI's economic diffusion, Microsoft's platform strategy, Xbox's sustainability challenges, encryption battles, AI-supercharged mass surveillance risks, and why privacy nihilism is a dangerous position in 2026.

Key Questions Answered

• •AI Economic Diffusion: Nadella argues the benchmark for meaningful AI impact is not model capability but whether marginal token costs match marginal productivity value across every enterprise. Token-maxing without measurable business returns does not drive GDP growth. Companies should audit AI spending against concrete output gains rather than deploying frontier models for non-frontier problems, using tiered model selection like Microsoft's Copilot auto-mode.
• •Agent-First Hardware Design: Microsoft's Project Solara envisions devices purpose-built for 24/7 agent workflows rather than retrofitting phones or PCs. Nadella describes a badge-sized device for workers like nurses that handles scanning, speech input, and prompt generation. Organizations planning AI infrastructure should consider whether current device form factors actually support persistent agent loops or merely replicate existing workflows with a new interface.
• •Xbox Sustainability Reset: Microsoft's Xbox division, now 25 years old, faces a structural monetization problem: more revenue from Xbox games flows through YouTube than through Microsoft itself. New Xbox leader Asha is conducting a 100-day review. Hardware component scarcity driven by AI and cloud demand is a temporary cost pressure, but the underlying business model question of how to capture entertainment value long-term is permanent.
• •Surveillance Business Model Shift: Cohn identifies the moment EFF stopped aligning with major tech platforms as when surveillance became their primary revenue model. Her framework for tech companies remains binary: stand with users and EFF stands with you; stand against users and EFF litigates. Organizations evaluating AI vendor partnerships should assess whether the vendor's revenue model depends on monetizing user data rather than delivering user value.
• •AI-Supercharged Mass Surveillance Risk: Cohn identifies AI-enhanced mass domestic surveillance as the single greatest privacy threat, noting Anthropic's own red lines prohibit it. The danger is not abstract: Nebraska prosecuted a woman using Facebook messages about reproductive healthcare. As surveillance capabilities expand, the population of people who become targets widens unpredictably, making privacy protections structurally important regardless of current political alignment or personal risk assessment.
• •Privacy Nihilism Is Strategically Wrong: Cohn counters the "data is already out there" fatalism by noting intelligence agencies treat old data as low-value because people continue generating new behavioral data daily. Cutting off mass data collection now degrades future surveillance capacity incrementally. The second-best time to pass comprehensive privacy legislation covering both commercial entities and law enforcement is today, not a decade ago.