Compose 2.5 by Cursor was trained at xAI's Colossus 2

@techdevnotes (Partly)

@techdevnotes cursor training on colossus is the part most people skipped reading. the data moat is dissolving into compute partnerships now. whoever owns the developer surface ends up renting the cycles from whoever owns the gpus. interesting decade ahead.

@techdevnotes Good let me try it...

Build a high-performance data center foundation with Supermicro infrastructure solutions. Optimized cabling design, routing, and port mapping help reduce time-to-online, while scalable water and dry cooling towers support thermal management and parallel expansion.

@techdevnotes This is a good news

@techdevnotes Very innovative

@techdevnotes Innovation at its peak🔥

@techdevnotes 10x compute on Colossus 2? The next model after Composer 2.5 is gonna be scary good. Excited to see what they cook up for long-context coding

@techdevnotes that explains the reasoning improvements

@techdevnotes 🎯

@techdevnotes And jjjina