Byte Points #114

This week on the pod, we dig into a growing paradox: is AI actually making us smarter or just faster and more burned out? Despite massive hype, new research shows productivity gains are narrow and uneven, with real improvements mostly limited to coding and customer support. Meanwhile, “AI brain fry” is emerging as a real phenomenon, with users experiencing cognitive fatigue, reduced critical thinking, and rising workloads instead of the efficiency they were promised. We also break down a major shift in AI security and power dynamics. Anthropic unveiled Project Glasswing alongside its unreleased Claude Mythos model—capable of autonomously discovering zero-day vulnerabilities across critical systems. It’s a double-edged sword: a powerful tool for defense, but one that could dramatically lower the barrier for cyberattacks if misused. That same tension shows up in the real world, where AI is now reportedly being used in military rescue operations, highlighting just how quickly these systems are moving from theory to high-stakes deployment. On the consumer side, AI continues to reshape everyday experiences. Google is pushing more offline-first AI with on-device dictation and interactive 3D simulations, while ElevenLabs enters the mobile music race with a generative AI app built for creation and discovery. But as convenience rises, so do risks—retailers like Target are already warning users they may be financially responsible for mistakes made by AI shopping agents acting on their behalf. In the enterprise world, Salesforce says AI has boosted productivity enough to pause engineering hiring entirely, while Microsoft continues expanding its Copilot ecosystem to the point of confusion—with nearly 80 different “Copilot” products now in circulation. At the same time, infrastructure pressures are intensifying: AI demand is reshaping cloud services, even forcing gaming platforms offline as compute gets reallocated to AI workloads. We also explore the growing backlash. Communities are pushing back against massive AI data centers over energy and water use, governments are questioning AI’s role in defense and regulation, and legal battles are heating up—with Anthropic now locked out of key Pentagon contracts amid national security concerns. Finally, we zoom out to markets and macro trends: inflation spikes tied to geopolitical conflict, continued volatility in stocks and crypto, and ongoing supply chain strain driven by AI’s appetite for chips and memory. We close with The Oracle: Anthropic quietly exploring custom AI chips as compute demand explodes—another signal that the next battleground in AI won’t just be models, but the infrastructure that powers them.