Chain of News 30/05/2026

**Top Story** A recent study, "How Coding Agents Fail Their Users: A Large-Scale Analysis of Developer-Agent Misalignment in 20,574 Real-World Sessions," has shed light on the limitations of AI coding agents in real-world scenarios. The study analyzed 20,574 coding sessions and found that existing analyses of agent failures often rely on benchmark trajectories that miss how developers actually experience misalignment. This research has significant implications for developers, as it highlights the need for more nuanced understanding of how coding agents interact with humans. The study's findings suggest that developers should be cautious when relying on AI coding agents, as they may not always align with human intentions. Furthermore, the study's results can inform the development of more effective coding agents that can better understand and respond to human needs. As AI coding agents become increasingly prevalent, this research serves as a reminder of the importance of careful evaluation and testing to ensure that these tools are truly beneficial to developers. **AI Models & Research** The Proprio framework, proposed in the paper "Proprio: Latent Self-Scoring and Inference-Time Refinement for Physically Plausible Video Generation," is a significant development in the field of video generation. Proprio enables a frozen video generator to assess and improve the physical plausibility of its outputs, resulting in more realistic and coherent videos. This research is important for developers, as it provides a new tool for generating high-quality video content. Another notable paper, "When Does Persona Prompting Actually Help? A Retrieval and Metric Analysis of Expert Role Injection in LLMs," investigates the effectiveness of persona prompting in large language models. The study finds that persona prompting can be beneficial in certain contexts, but its practical value remains unclear. The ClothTransformer, presented in the paper "ClothTransformer: Unified Latent-Space Transformers for Scalable Cloth Simulation," is a unified and scalable transformer model for cloth simulation, which has achieved remarkable success in modeling diverse phenomena. These research developments have the potential to significantly impact the field of AI and machine learning, and developers should be aware of these advancements to stay up-to-date with the latest technologies. **Developer Tools & Frameworks** Microsoft has announced the general availability of Agentic Coding tools, which allow developers to build Power Pages sites with AI. This update enables developers to leverage the power of AI to create more efficient and effective coding workflows. With Agentic Coding tools, developers can now use AI to automate repetitive tasks and focus on more complex and creative aspects of coding. Additionally, the DynoSim framework, presented in the paper "DynoSim: Simulating the Pareto Frontier," provides a new approach to simulating the Pareto frontier in modern LLM serving. This framework allows developers to better tune their LLM deployments and optimize their performance. By leveraging these new tools and frameworks, developers can improve their productivity and create more innovative solutions. **Industry & Business** Microsoft is set to unveil its in-house AI models for coding, reasoning, and images at the Build conference. This announcement is significant, as it highlights Microsoft's commitment to developing and deploying AI technologies. The company's in-house AI models have the potential to revolutionize the way developers work and create new opportunities for innovation. In another development, Microsoft has made its Agentic Coding tools generally available, allowing developers to build Power Pages sites with AI. This move demonstrates Microsoft's efforts to democratize access to AI technologies and provide developers with the tools they need to succeed. These developments are likely to have a significant impact on the industry, and developers should be aware of these changes to stay ahead of the curve. **Worth Watching** The use of AI to measure the speed of brain cleaning during sleep, as reported in the article "Nuevo hito contra el Alzheimer | La inteligencia artificial logra medir la velocidad de limpieza del cerebro durante el sueño," is a fascinating development that deserves attention. This research has significant implications for our understanding of Alzheimer's disease and the potential role of AI in diagnosing and treating neurological disorders. Another interesting development is the Annotator Positionality as Signal approach, presented in the paper "Annotator Positionality as Signal: Psychometric Weighting for Anti-Autistic Ableism Detection." This research highlights the importance of considering annotator positionality in large language models and has significant implications for the development of more inclusive and equitable AI systems. These developments are worth watching, as they have the potential to significantly impact the field of AI and machine learning.