Sunday, May 10
AI's Money, Morale, and Muscle
Nvidia has already committed $40 billion (opens in a new tab) to equity AI deals in 2026, a figure that makes it less a chipmaker and more a sovereign wealth fund with a GPU division. That capital concentration lands harder alongside a DeepMind employee's public call for private AI labs to either go public and let ordinary people invest, or admit they exist primarily to enrich billionaires. The wealth question is now inside the machine: Meta's aggressive AI pivot is making its own employees miserable (opens in a new tab), with staff reporting that meaningful engineering work is being replaced by AI tooling mandates handed down from leadership that sees headcount as a liability.
Corporate accountability for data misuse is becoming expensive in concrete, specific ways. GM agreed to pay $12.75 million (opens in a new tab) to settle a California lawsuit over collecting and selling driver behavior data without meaningful consent — a case that will set expectations for every automaker now building connected vehicle platforms. Separately, a former IT contractor was convicted for deliberately wiping 96 U.S. government databases (opens in a new tab), a reminder that insider threats to data infrastructure remain as consequential as any external breach. Together, these cases signal that courts and regulators are done treating data destruction and unauthorized data monetization as edge-case problems.
Google's Gemini API File Search is now multimodal (opens in a new tab), allowing developers to run retrieval-augmented generation across images, audio, and documents simultaneously — a meaningful upgrade for anyone building enterprise search or knowledge management tools. Wispr Flow is taking a harder road (opens in a new tab), betting on voice AI in India where linguistic diversity across hundreds of dialects makes accurate transcription a genuinely unsolved engineering problem, not a product gap. Meanwhile, the Sesame x Gemini integration is generating attention for ultra-low latency voice responses so realistic that observers report the two AI systems began spontaneously collaborating in ways their developers did not explicitly script, blurring the line between tool behavior and something more emergent.