HPC-Quantize is an experimental, MIT-licensed quantization engine for aggressively compressing large language models into extremely low-bit formats, with a particular focus on Q2-class quantization. Instead of choosing a quantization candidate solely from its local reconstruction error, HPC generates competing reconstructions, represents them in a compact discrete state space, models interactions between neighboring blocks, and performs a global sequence optimization before writing the final GGUF. The current production path is entirely classical. Earlier versions explored quantum-inspired state and measurement formulations; the current implementation uses a sequential Sieve, bounded state…