Can AI design circuit boards yet? (eebench.org)

OpenAI recently showcased a demo of its GPT-6 Astra model effectively working on circuit board designs within the KiCad environment, sparking excitement in the AI/ML community. This demonstration highlights significant advancements in how AI can better understand and participate in electronics design, moving beyond theoretical knowledge to practical applications. While we are still far from AI autonomously producing complex designs, the potential for AI to assist in designing and simulating real-world circuitry is becoming evident. A new tool called EEBench offers a more structured approach to evaluating AI's capabilities in electronics by using declarative code instead of conventional graphical CAD tools. This allows AI models to directly manipulate circuit components and constraints, significantly improving the design process. EEBench conducts rigorous simulations, measuring performance against real-world specifications and tolerances, thus enabling a more accurate assessment of an AI's engineering proficiency. Recent results show promising scores for models like Claude Opus 5 and Grok 4.6, indicating substantial progress in AI-driven electronics design. As AI labs start to take electronics seriously, the field is poised for further developments that could eventually lead to safer and more reliable hardware designs.