Fresh articles on Habr about artificial intelligence (April 19): 1. "AI Development Strategy in Russia: How Strategies Are Written When the Planning Horizon Is No More Than a Year" - a critical analysis of systemic errors in innovation management and corruption risks in state AI strategies. The author examines the problems of short-term planning in a rapidly changing field. 2. "How CMA-ES Works for Hyperparameter Optimization in Optuna" - a technical article about the evolutionary strategy method for optimizing complex functions without gradients. Useful for ML engineers working with model tuning. 3. "How AI Sycophants Drag You into AI Psychosis, or What Toxic Agreement Leads To" - from MTS about the problem of AI hallucinations and toxic agreement in corporate environments. Examines the risks of blind trust in neural network conclusions. 4. "How I Became a Vibe Coder (Though I Probably Seemed Like a Respectable Person)" - the author's personal experience about transitioning to new programming approaches. Trends: Interest remains in critical analysis of AI strategies, technical optimization methods, and issues of trust in neural networks. The first article may be interesting for outreach to experts in public administration and innovation.