I love deepseek_ai v4.1-flash even more today.
I don't need to use Opus-5.5 anymore. Well for short presentation type videos you don't need to spent much time working on graphics anymore. This entire 60s video on KV cache and context length is a motion graphic animation
The US blocked NVIDIA's best chips from China.
Chinese universities didn't stop building AI servers. They just changed the parts list. This one runs DeepSeek V4 for a uni in Shanghai: • GPUs: NVIDIA L20, a card NVIDIA made only for China so it passes export rules • RAM: CXMT.
I deleted my ChatGPT subscription.
Ollama runs DeepSeek and Llama on my own laptop. One command. Works in airplane mode. Zero API bills, total privacy.
I used Deepseek to generate popular science videos based on the methodology used by Opus 5.5 to generate videos.
"Historical Evolution of Medical Methodology"
You have reached the end of the archive
All of deepseek