2026-06-08

1. Apple WWDC 2026 — iOS 27 Multi-Model AI Chooser: Claude + ChatGPT + Gemini at OS Level

• Apple unveiled Siri 2.0 powered by a custom 1.2-trillion-parameter Gemini model licensed from Google DeepMind for ~$1B/year, running in Apple's Private Cloud Compute (no data retention).
• iOS 27 / iPadOS 27 / macOS 27 introduce a system-wide "Search or Ask" panel (swipe down from top) with a user-selectable AI backend: Siri (Gemini), Claude (Anthropic), or ChatGPT (OpenAI).
• Apple becomes the first company to integrate all three frontier AI providers simultaneously at the OS level, reaching ~1.4 billion active devices.
• Why it matters: The consumer AI battle shifts to the platform layer. Google wins the largest commercial AI licensing deal ($1B/year, 1.4B devices). Anthropic and OpenAI gain first-party OS-level presence without launching dedicated consumer apps. Apple avoids frontier-model competition by becoming a neutral router.
• → Apple (new) · Google DeepMind · Anthropic · OpenAI
• (source) (MacRumors) (Enterprise DNA)

2. NVIDIA EgoScale — Dexterous Humanoid Trained from 20K Hours of Human Video (No Robot in Loop)

• Jim Fan (@DrJimFan, NVIDIA GR00T co-lead) announced that his team trained a humanoid with 22-DoF dexterous hands to assemble model cars, operate syringes, sort cards, and fold shirts — using 20,000+ hours of egocentric human video with no robot teleoperation.
• Key scientific result: log-linear scaling law (R² = 0.998) between human video volume and action prediction loss, which directly predicts real-robot task success rate.
• Everything open-sourced: weights, code, post-training dataset, eval set, and EgoScale whitepaper.
• Additional demo: live VR teleoperation of a virtual robot on Unitree G1 via PICO headset.
• Why it matters: Confirms that the LLM scaling paradigm (more data → predictably better performance) transfers directly to physical dexterous manipulation. The R²=0.998 scaling law is the strongest evidence yet that egocentric human video is a viable foundation for general-purpose robot training — a potential turning point in robot data strategy.
• → Jim Fan · Embodied Agents · NVIDIA
• (source) (X post) (Humanoids Daily)