Vitalik Buterin has outlined a plan for deploying localized, private large language models (LLMs) with a focus on privacy and security. The initiative aims to minimize the risk of data leakage and unauthorized access by avoiding remote models and external services. Key strategies include local inference, on-device file storage, and sandbox isolation.Buterin's hardware tests involved a laptop with an NVIDIA 5090 GPU, an AMD Ryzen AI Max Pro device with 128 GB unified memory, and the DGX Spark. Performance results showed the 5090 laptop achieving 90 tokens per second with the Qwen3.5 35B model, while the AMD device and DGX Spark reached 51 and 60 tokens per second, respectively. Buterin favors high-performance laptops for building local AI environments, utilizing tools like llama-server, llama-swap, and NixOS.