SignalChina/HK

DSV4-Pro is really good at Kernel, this is promising sign for SWE & RSI. At this pt, the question is just how fast DeepSeek team can iterate & improve themselves, bc both Kimi & GLM models came out ea

What changed

I cheer for DS bc it has the most efficient approach. That's the only way we can get the cost of AI down even more. — via @tphuang

Why it matters

DeepSeek's efficiency and rapid iteration signal SWE/RSI gains, but lacks hard data.

Sources