As I watch the 99% cache hit rate, 200+ tps & running additional "workers" to speed up execution of my task. I also reflect on other ppl experience.

What happened: A post describes observing a 99% cache hit rate and 200+ tps while running additional workers to speed up task execution, and reflects on others' experience, including what it calls Qoder's failed effort to make Qwen appear to be DeepSeek Flash. Key numbers: 99% cache hit rate; 200+ tps. Why it matters: The post argues people use DS because it is good enough, fast and cheap, a combination it says is the only one able to serve China's growing AI work force without major bottlenecks, which it says lets WorkBuddy serve more customers than Qwen Work.

A post describes observing a 99% cache hit rate and 200+ tps while running additional workers to speed up task execution, and reflects on others' experience, including what it calls Qoder's failed effort to make Qwen appear to be DeepSeek Flash.

The post argues people use DS because it is good enough, fast and cheap, a combination it says is the only one able to serve China's growing AI work force without major bottlenecks, which it says lets WorkBuddy serve more customers than Qwen Work.