Ahahahaha. Liang Wenfeng and the Blue Whale anthem is exactly what I didn't know I need related to DeepSeek!

APT APT ! 😂😂

Daebak 앤디Oct 41

Qwen, DeepSeek and Mistral can turn that expensive GPU into the employee who takes the night shift

I turned the questions into an interview and let one tiny spider steal the useful words VRAM → leave room for the conversation Q4 → squeeze the weights, check the answers

goguOct 49

America's Treasury Secretary just accused China's most powerful AI model of being stolen, down to the personality.

Bessent told Axios that Chinese labs practice "industrial distillation," which he called "a nice word for stealing from the U.S. models." His example: Moonshot's

Nathan PellerinOct 41

One session of a coding agent: 152 turns, 1,097 steps, 945 tool calls

Every single one of them run_code. The other 19 tools all sit behind that one name. Made a film about that shape, with fframes.