fuck you sam altman. i am keeping my spare capacity.
Update from the lawyer with the V100 server. Some of you have seen the earlier posts. The short version: one box of the pile now runs DeepSeek-V4.1-Flash, a big open model, with every weight in GPU memory and a 524K-token context, and it does real work for me every day.
What it's for
Private first-pass work on my own case files. The documents never leave the office. It reads a folder of records, pulls out the facts, dates and numbers, summarizes, and takes a first swing at routine drafting. Nothing goes out the door without a real review. By my estimate 50-60% of my existing LLM workflows can now run on this box instead of frontier models, at speeds that match Claude Code in the chat window. My RAG databases and tool calls are wired in. I drive it in plain English through a Discord-style interface over SSH from my Windows PC.…