Geometric deep learning for materials discovery in 2026 — GNoME, MatterGen, equivariant GNNs, the inverse design problem, and where ML-driven materials science actually ships.
How atomic clocks actually work — cesium hyperfine transitions, optical lattice clocks, what makes them so accurate, and why GPS and the internet depend on them.
How DeepMind AlphaProteo designs de novo protein binders — diffusion + structure prediction, target-conditioning, in vitro hit rates, and the 2026 outlook for AI-designed therapeutics.
AlphaFold 3 architecture explained: the Pairformer, the diffusion module, how it predicts proteins, DNA, RNA, ligands and complexes, and its accuracy and limits in 2026.