On August 5, Jeff Dean announced his departure from Google, where he had worked for 27 years, to found a company called Discovery Loop, aiming to enable AI to enter a continuously running scientific discovery loop. A Google DeepMind paper titled *LLM can't jump* also went viral, referring to a leap in scientific discovery that Einstein mentioned, where scientists form new basic principles from empirical facts through a leap not fully derivable by existing logic. Zahavy's judgment suggests models may automate experiments and complete complex mathematical deductions but lack the critical step of inventing previously non-existent basic principles. Zahavy disagrees that compression covers the whole process, stating it is only one of three modes for forming knowledge: induction, deduction, and abduction, which is inventing new explanatory principles from limited facts. The author views current models as structurally incapable of abduction, as Einstein's approach relied on translating simulated experiences into formal axioms through thought experiments, while current LLMs mainly process symbolized content and lack an action-controllable world model to complete those simulations.