Week 3 of July 2024 saw global AI advances in reasoning, on-device AI, ethics, and competitive dynamics. Key developments: OpenAI developing "Strawberry" (codename) with human-level reasoning approaches — using Self-Taught Reasoner (STaR) method for iterative model retraining to progressively improve reasoning; specifically targeting mathematical problem-solving beyond current AI limits. Hugging Face released SmolLM (ultra-lightweight model running on mobile devices without internet connection) for on-device AI democratization. OpenAI faced controversy over allegedly bypassing internal safety testing procedures for GPT-4o launch and SEC investigation requests. EU''s AI Act entered into force August 1, strengthening facial recognition and general-purpose AI system regulations, mandating copyright disclosure for AI training data. Anthropic launched $100M Anthology Fund for AI startup support. UK government AI Safety Institute published first international AI safety evaluation framework. Stability AI raised new funding round for Stable Diffusion development continuation. Key insight: the gap between AI capability advancement speed and regulatory/governance framework development is widening — the EU AI Act represents the most comprehensive governance attempt but was designed before the current generation of frontier models, creating questions about whether rules designed for previous-generation systems adequately address emerging risks.