Does Scaling Toward AGI Depend on What System-1 Lets System-2 Execute?

AI has grown by paying for more: more parameters, more data, longer reasoning, more agents. In a new Perspective from Tsinghua University, we argue the next gains may come from somewhere else:

AI has grown by paying for more: more parameters, more data, longer reasoning, more agents. In a new Perspective from Tsinghua University, we argue the next gains may come from somewhere else: deciding, cheaply and before paying, which of those interactions are worth it. Paper: arxiv.org/pdf/2610.39497 Project website, with an interactive demo: cheemswang.github.io/blogs/sys1-sys2-codesign.html Qi Wang, Yun Qu, Lizhou Cai, Heming Zou, Haoang Chi, Yiqin Lv, Xiang Li, Haiyan Feng, Mengyu Ge, Yize