Day 35 of testing DGX Spark with DeepSeek Harness & Qwen 3.8.
Encountered setup issues, power loss, and slow token speeds (10-12 tokens/sec vs. expected 60). Memory usage is high (90GB). Long processing times (88 mins) suggest potential issues. Will revisit tomorrow.
NVIDIA is opening APIs for a series of extremely powerful AI models.
Just need 1 API key to test DeepSeek, MiniMax, Qwen, Kimi, GLM and many other models via OpenAI compatible endpoint. If you're building an Agent, chatbot or AI app, it's worth testing. Save this video or follow
The most compelling version of why the AI bubble might burst isn't about the models.
It's about electricity. AI data centers are adding power capacity faster than any industry has ever built generation capacity in U.S. history. The grid interconnection queue sits at 474
One day before that disaster I had a devastating dream and how I keep everything
Watch the video first Deepseek gets confused and I ask him the date again... I'm not a psychic, I think that sometimes the universe sends you signs, but what's the point if you don't even know precisely where it is or will be.
You have reached the end of the archive
All of deepseek