Yao Shunyu, head of Tencent Hunyuan, revealed the real moat of large AI models in one sentence!
I was asked by a professor from Tsinghua University about DeepSeek, Claude, GPT and Hunyuan Hy, which one can retain users better? He said directly: The first moat of the model is not the parameters, but the rare context; the second is whether you can trust it to control your data. Trust + unique scene memory is the real killer of user stickiness.
Qwen4 built me an explorable pokemon voxel world.🐧
Honestly DeepSeek 1017 did a much better job overall with this but i do appreciate the pokeball throw feature it came up with. click a pokemon, ball arcs in, shakes, gotcha.
Is there an AI bubble brewing?
Hamid Rashid, executive chairman of Global AI, says the US has seen extensive overinvestment in AI infrastructure, with returns that may be difficult to justify. What about China? He sees a different picture, pointing to models like DeepSeek that
You have reached the end of the archive
All of deepseek