As Rocks May Think (evjang.com)

A significant update in AI capabilities has emerged, showcasing advanced coding and reasoning abilities among machines. Since 2022, developments like ChatGPT and Claude Code have fundamentally changed how researchers and developers interact with AI, allowing for the automation of tasks like code generation and experimental design. The article details the author's experiences using Claude to create an "automated AlphaGo researcher," capable of performing intricate tasks such as hyperparameter optimization and generating experiment scripts autonomously. This evolution marks a shift from traditional automated systems, which were limited to predefined spaces, to modern coding agents that can reflect on results and adapt methods dynamically. The implications of these advancements are profound for the AI/ML community. Modern coding agents can now tackle a vast array of digital problems, potentially revolutionizing fields ranging from software development to scientific research. With the ability to execute complex experiments and formulate hypotheses, these AI systems are stepping into the roles of researchers and software engineers, thereby accelerating technological progress. This capability suggests that we are entering a "golden age" where AI can help address previously infeasible challenges in computation, opening new frontiers in areas like physics discovery and algorithm design. The rapid evolution of these systems points to a future where coding assistants become essential, enabling engineers to effortlessly create intricate applications and solve complex computational issues.