I started a project 10 years ago to build my own #memex and now that I can train it with my own corpus based on precedent set by #Anthropic I'm 100% sure I can make better models than $10 billion companies and I'll have the hardware to do it in 2, developing my own tools on commodity hardware.