The recent advancements in the research area of large language models (LLMs) and blockchain naming systems (BNS) have shown significant strides in addressing critical challenges such as privacy, efficiency, and security. In the realm of LLMs, there is a notable shift towards optimizing inference processes to maximize throughput while minimizing latency, particularly through innovative CPU offloading techniques and speculative decoding methods. These approaches aim to alleviate the constraints imposed by limited GPU memory and enhance the overall performance of online LLM services. Additionally, the integration of privacy-preserving frameworks is becoming increasingly important, ensuring that user data remains anonymous and secure against potential exploitation by service providers. On the blockchain front, the focus has been on mitigating typosquatting attacks within BNS, which pose a significant risk to user funds. Researchers are proposing straightforward countermeasures to protect users from such malicious activities. Furthermore, advancements in serverless computing and GPU-oriented data transfer are being explored to optimize data handling and reduce latency in machine learning inference applications. These developments collectively underscore a trend towards more efficient, secure, and privacy-conscious technologies in both LLM and blockchain domains.