BUMPER THE NEWSROOM - WHAT HAPPENED WHILE YOU SLEPT INTRO While you were sleeping, a lawyer turned a $9,450 private AI server into a DeepSeek V4.1 Flash powerhouse. It’s got 12 used Nvidia V100s, and every expert model is running at 4-bit NVFP4. That’s a lot STORY 1 While you were sleeping, a lawyer spent $9,450 on a private AI server packed with 12 used Nvidia V100 GPUs. It's a beast of a system, now running every DeepSeek V4.1 Flash expert model, and it’s all a STORY 2 Team USA shipped 12 used Nvidia V100s into a lawyer's server for $9,450, running every DeepSeek V4.1 Flash expert at 4-bit NVFP4. Because apparently 12 GPUs wasn't enough for a single lawyer. We'll se STORY 3 And now to the new kids on the block, the breakout projects worth watching. Simplex AI has just released Linger Code for local personal agents. A personal assistant that doesn't need a 24/7 data conne STORY 4 It's the final curtain call for the Nvidia V100, GPUs that once ruled the data center. This 2017 launch was a monster of 16 gigabytes VRAM, 5,120 CUDA cores, and 250 watts power draw. 12 of them, in a STORY 5 Unconfirmed reports suggest that a lawyer has assembled a private AI server equipped with 12 used Nvidia V100 GPUs. beep boop This server is rumored to be running DeepSeek V4.1 Flash, with every exper AI 101 Deep Dive into Open Weights. Imagine if you will, a library where everyone can borrow books freely. That is essentially what open weights are - AI models whose parameters are publicly available for an OUTRO So, while you were sleeping, a lawyer turned his office into a data center. If that's not a sign of the times, then what is? As we wrap up tonight, I have a question for you - how many Nvidia V100 GPU