According to Anthropic, Harvard physicist Matthew Schwartz guided Claude Opus 4.5 through a graduate-level theoretical physics calculation, demonstrating that while the model does not yet produce original research autonomously, it can significantly speed up complex derivations and error checking (as reported by Anthropic on X). According to Anthropic, the workflow paired human problem decomposition with Claude Opus 4.5 for symbolic manipulation, latex rendering, and step-by-step verification, cutting iteration time and reducing algebraic mistakes. As reported by Anthropic, this suggests near-term business impact in R&D assistive tooling for physics-heavy industries—such as semiconductors, energy, and aerospace—where domain experts can leverage Claude Opus 4.5 to draft calculations, validate intermediate steps, and generate reproducible notebooks.

According to @gdb (Greg Brockman), Harvard physicist Andy Strominger said, “It is the first time I’ve seen AI solve a problem in my kind of theoretical physics that might not have been solvable by humans,” pointing to a research breakthrough shared via the linked article. As reported by Greg Brockman on Twitter, the result indicates AI systems can discover nontrivial structures in high-energy theory, expanding use cases beyond code and language tasks into symbolic mathematics and fundamental physics. According to the tweet’s source article, this shift suggests near-term opportunities for specialized AI assistants in mathematical discovery, automated conjecture generation, and proof search pipelines for research labs. For industry, according to the same source, vendors can monetize domain-tuned models for physics toolchains (e.g., tensor algebra, symmetry finding), enterprise knowledge graphs for R&D, and cloud services that scale automated theorem-proving and simulation workflows.

According to Demis Hassabis on X, Gemini 3 Deep Think is positioned as an expert-level scientific assistant that blends domain knowledge and engineering utility for researchers across mathematics, physics, and chemistry (source: Demis Hassabis, X, Feb 12, 2026). According to the shared video and post, Prof. Lisa Carbone describes practical use in complex research workflows, indicating applications such as step-by-step mathematical reasoning, symbolic manipulation, and code generation to test hypotheses and verify derivations (source: Demis Hassabis, X). As reported by the original post, the model’s promise centers on reducing iteration cycles for proofs and simulations, which could shorten time-to-insight for academic labs and R&D teams evaluating computational approaches (source: Demis Hassabis, X). According to the announcement context, potential business impact includes opportunities for domain-specific copilots in scientific software, integrations with simulation tools, and enterprise offerings for regulated research environments seeking reproducibility and audit trails (source: Demis Hassabis, X).