In a guest contribution to Terence Tao’s blog, Francesco Maggi argues that more mathematical results from AI do not necessarily mean growth in human understanding or an independent research base. He distinguishes checking long proofs with Lean from understanding new concepts and building theories, and raises questions about assessing student contributions and protecting unpublished research. Proposed responses include assured access, disclosure of use, reproducibility standards, and public research infrastructure. This is Maggi’s commentary, not Tao’s own authorship or experimental evidence establishing AI’s educational effects.