[from] 李继刚
Daily Paper Reading：
Why haven't models gotten much bigger in the past two years, yet Agents suddenly started getting stuff done?
为什么过去两年模型规模没怎么增长，智能体（Agents）却突然开始能干活了？
The story of the past three years is a familiar one: parameters kept piling up, from billions to trillions, with GPT-4, Gemini, DeepSeek building layer upon layer. Everyone believed in the scaling law, believed that as long as the model was big enough, it could remember the world, plan long tasks, and collaborate with others. This was the era of "solving problems with sheer mass."
