The field of software and hardware development is witnessing a significant shift towards AI-driven optimization, with a focus on improving performance, reducing costs, and enhancing interpretability. Recent research has explored the potential of large language models (LLMs) in automating tasks such as code generation, optimization, and design space exploration. Notably, LLMs are being used to generate optimized CUDA code, refine GPU kernels, and assist in architectural design space exploration. Moreover, researchers are investigating ways to improve the efficiency and effectiveness of LLMs, such as through the use of multimodal assistants, turn-control strategies, and retrieval-augmented generation frameworks. These advancements have the potential to revolutionize the field of software and hardware development, enabling faster, more efficient, and more effective development of complex systems. Noteworthy papers include: Multimodal Chip Physical Design Engineer Assistant, which introduces a multimodal large language model assistant for physical design optimization. STARK: Strategic Team of Agents for Refining Kernels, which presents an LLM agentic framework for GPU kernel optimization. From Large to Small: Transferring CUDA Optimization Expertise via Reasoning Graph, which proposes a training-free, retrieval-augmented generation framework for transferring LLM-level reasoning to smaller models.