Designing customized processors is a complex and time-consuming endeavor. Traditionally, it requires extensive manual effort and expertise in hardware description languages (HDLs) like Verilog or Chisel. Imagine, however, if you could simply describe the desired functionality of a processor in plain English and have an AI automatically generate the complex RTL code needed to bring it to life. That’s the promise of AGON, a groundbreaking new framework that leverages the power of large language models (LLMs) to automate the design of customized processors. The key innovation behind AGON is its introduction of “nano-operator functions” or nOP functions. These functions act as a bridge between high-level descriptions of instruction sets and the nitty-gritty details of hardware implementation. Think of nOPs as building blocks that represent fundamental hardware operations. By combining these blocks, AGON can describe complex instructions in a way that's both efficient and easy for LLMs to understand. This streamlined approach allows AGON to bypass many of the challenges that have plagued previous attempts at AI-driven hardware design, such as the difficulty LLMs have with managing the scale and complexity of HDL code. Furthermore, AGON cleverly decouples the functional description of a processor from the task of optimizing its performance, power, and area (PPA). This separation of concerns allows the LLM to focus solely on generating code that correctly implements the desired functionality. AGON then takes over, using a set of sophisticated primitives to optimize the generated design at the instruction, ISA, and processor levels. This process involves techniques like merging redundant operations, fusing multiple nano-operations into single, more efficient units, and exploring different configurations to find the best balance between performance, area, and power consumption. Early experiments with AGON have shown promising results. It has successfully designed a series of customized out-of-order processors for various applications, achieving a significant performance boost compared to a standard, expert-designed general-purpose CPU. In some cases, AGON-designed processors ran specific tasks up to 17 times faster. These initial successes suggest a paradigm shift in processor design is within reach. AGON has the potential to democratize chip development, allowing a broader range of developers to create customized hardware tailored to specific needs. While still in its early stages, AGON represents an exciting step towards a future where AI plays a central role in creating the next generation of computer processors. This will significantly accelerate the design process and unlock new possibilities for innovation in areas like embedded systems, artificial intelligence, and beyond. However, challenges remain in extending the framework to support a wider range of processor architectures and incorporating more advanced optimization techniques. The ongoing research around AGON promises to refine its capabilities and pave the way for truly autonomous chip design.