Essay 01 · in Chinese
为什么 LLM 不能做自主科学发现，世界模型亦然
Why LLMs Cannot Discover, and Neither Can World Models
The computational axis (parallel, copyable, scalable, δ = 0) and the temporal axis
(unrepeatable, irreversible, mortal, δ > 0) are not orthogonal — they exclude each other.
Learning is θ-change; discovery is Φ-change, and no amount of scaling crosses the break.
The Endogenous Discovery Theorem, the Law of Discovery, the anesthetization of civilization,
and the narrow path: artificial suffering as the only known engine of ASI.
Yilei Shao2026.9.8 · 2nd draftRead →